From rogerdpack at gmail.com Wed Mar 4 13:19:33 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Wed, 4 Mar 2009 11:19:33 -0700 Subject: [Rubyinstaller-devel] dsl branch In-Reply-To: <71166b3b0902251928h4caf0ba5t9d9dbcfbfa7552e4@mail.gmail.com> References: <71166b3b0902251928h4caf0ba5t9d9dbcfbfa7552e4@mail.gmail.com> Message-ID: > Is a working in progress and I need to push the latest changes I have locally. Sounds good. As a note I did see a few mingw related recipes at http://github.com/janneke/gub/tree/d300719be162c60d12007369117f8bfd0c85ed1f/gub/specs I almost wonder if there's not a way to hook into cygwin's repositories to create a more generic build system. Like http://donc.wordpress.com/2006/08/04/apt-for-cygwin/ or http://stephenjungels.com/jungels.net/projects/apt-cyg/ But then again, I suppose we already know how to build the dependencies [except tcl/tk -- any volunteers to add them?] so our recipes are probably good enough. Thoughts? -=r From rogerdpack at gmail.com Sat Mar 14 11:45:10 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Sat, 14 Mar 2009 09:45:10 -0600 Subject: [Rubyinstaller-devel] md5 Message-ID: Currently it appears the mingw create script skips downloads if they already exist on the file system. Maybe we could add md5 sums for each file so we can check the downloads? -=r From luislavena at gmail.com Sat Mar 14 12:09:30 2009 From: luislavena at gmail.com (Luis Lavena) Date: Sat, 14 Mar 2009 12:09:30 -0400 Subject: [Rubyinstaller-devel] md5 In-Reply-To: References: Message-ID: <71166b3b0903140909u2abf036evac87d0c045c962f1@mail.gmail.com> On Sat, Mar 14, 2009 at 11:45 AM, Roger Pack wrote: > Currently it appears the mingw create script skips downloads if they > already exist on the file system. ?Maybe we could add md5 sums for > each file so we can check the downloads? Downloads are placed in a temporary folder while download. once transfer is complete, they are moved to the final location. Since is using streaming (and not open-uri) it should not have discrepancies in the actual files being downloaded. But I guess that having md5 hashes could be helpful. Keep in mind that some of the downloads are 4MB, if md5 computation needs to happen on those files, needs to be faster. -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From rogerdpack at gmail.com Sat Mar 14 23:10:19 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Sat, 14 Mar 2009 21:10:19 -0600 Subject: [Rubyinstaller-devel] md5 In-Reply-To: <71166b3b0903140909u2abf036evac87d0c045c962f1@mail.gmail.com> References: <71166b3b0903140909u2abf036evac87d0c045c962f1@mail.gmail.com> Message-ID: > Downloads are placed in a temporary folder while download. once > transfer is complete, they are moved to the final location. That's probably good enough :) As a note I still get this message at compile time Creating library file: libmsvcrt-ruby18.dll.a -e:1: syntax error, unexpected tSTRING_BEG, expecting $end (I noticed Luis got it at least once before [1]). I think it's related to the weird "-e quoted strings don't work" Thankfully it doesn't seem to hurt anything so I guess who cares. As another note, I sometimes run into permissions issues when I 'overwrite' any files. Like do a successfull rake, then run rake xxx [compile or what not]. Any thoughts? Also a question. If I search for luislavena-mysql-gem it should appear shouldn't it? c:\dev\downloads\substruct_rel_1-0-a6>gem search mysql --source http://gems.github.com -b *** LOCAL GEMS *** mysql (2.7.3) *** REMOTE GEMS *** eejjjj82-mysql-ruby (2.8.1) elia-mysql (2.8) jonbrenner-mysql (2.8) kwatch-mysql-ruby (2.8.1) methodmissing-mysqlplus_adapter (1.0.4) mplatov-mysql (2.8.0) neovintage-dm-mysql (0.0.2) oldmoe-mysqlplus (0.1.0) oldmoe-neverblock-mysql (0.1.0) tmm1-em-mysql (0.2.0) seems missing for some reason? any thoughts on this? Also to get it to work do you need to install mysql from source [via mingw]? Anybody know if it works with 5.1? Thanks! -=r [1] http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/21903 From rogerdpack at gmail.com Sun Mar 15 00:56:57 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Sat, 14 Mar 2009 22:56:57 -0600 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <49BC09D8.8010604@savagexi.com> References: <49AF7F4C.1020809@savagexi.com> <49B7F268.50309@savagexi.com> <49B84F36.9060301@savagexi.com> <49B880F2.9060203@savagexi.com> <49BC09D8.8010604@savagexi.com> Message-ID: Perhaps we should [regardless of whether naming changes are approved or not] by default link against msvcr90 [from the mingw builds], and also include the VC2008 runtimes. That blesses the lives of the VC2008 people, since they don't have to worry quite as much about differing runtime library incompats. And it probably wouldn't be too hard to link against it from mingw. Is it possible using the cross compiler tho? Thoughts? -=r On Sat, Mar 14, 2009 at 1:47 PM, Charlie Savage wrote: > > > Roger Pack wrote: >> >> http://www.gamedev.net/community/forums/topic.asp?topic_id=482230 >> that type of thing won't be a problem, right? or is that the same >> problem as described in your previous link to >> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36834 > > That's a good question. ?Seems like it could the same issue since they > mention stack corruption. ?If they had posted all the source code we might > be able to tell - I'm not sure there is enough info there to tell. > > Charlie > > From luislavena at gmail.com Sun Mar 15 14:40:13 2009 From: luislavena at gmail.com (Luis Lavena) Date: Sun, 15 Mar 2009 14:40:13 -0400 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: References: <49AF7F4C.1020809@savagexi.com> <49B7F268.50309@savagexi.com> <49B84F36.9060301@savagexi.com> <49B880F2.9060203@savagexi.com> <49BC09D8.8010604@savagexi.com> Message-ID: <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> On Sun, Mar 15, 2009 at 12:56 AM, Roger Pack wrote: > Perhaps we should [regardless of whether naming changes are approved > or not] by default link against msvcr90 [from the mingw builds], and > also include the VC2008 runtimes. ?That blesses the lives of the > VC2008 people, since they don't have to worry quite as much about > differing runtime library incompats. ?And it probably wouldn't be too > hard to link against it from mingw. ?Is it possible using the cross > compiler tho? > Thoughts? > -=r > AFAIK, none of the mingw32 packages for the different linux distributions bundle msvcr90 libraries. Without those, after installing (macport or apt-get package) manually find/build/install the msvcr90 libraries to properly link to the specified runtime. After that, if the user wanted to execute using Wine (to run specs/test and verify the build doesn't segfault) they will need to follow the winetricks: http://wiki.winehq.org/FAQ#head-a0aa6667e9ba22fad996f7f0d51dc7d161087611 Another issue I see with this, I tell ruby configure about it, but there is a MSVCRT chooser script that can fake/hide the trick for us. I'm starting to think that is we are letting MSVCR90 be the default runtime, why I should bother with MinGW at all. I should be building Ruby with VC2008 instead... -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From rogerdpack at gmail.com Mon Mar 16 07:40:43 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Mon, 16 Mar 2009 05:40:43 -0600 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <49B84F36.9060301@savagexi.com> <49B880F2.9060203@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> Message-ID: On Sun, Mar 15, 2009 at 12:40 PM, Luis Lavena wrote: > On Sun, Mar 15, 2009 at 12:56 AM, Roger Pack wrote: >> Perhaps we should [regardless of whether naming changes are approved >> or not] by default link against msvcr90 [from the mingw builds], and >> also include the VC2008 runtimes. ?That blesses the lives of the >> VC2008 people, since they don't have to worry quite as much about >> differing runtime library incompats. ?And it probably wouldn't be too >> hard to link against it from mingw. ?Is it possible using the cross >> compiler tho? >> Thoughts? >> -=r >> > > AFAIK, none of the mingw32 packages for the different linux > distributions bundle msvcr90 libraries. > > Without those, after installing (macport or apt-get package) manually > find/build/install the msvcr90 libraries to properly link to the > specified runtime. > > After that, if the user wanted to execute using Wine (to run > specs/test and verify the build doesn't segfault) they will need to > follow the winetricks: > > http://wiki.winehq.org/FAQ#head-a0aa6667e9ba22fad996f7f0d51dc7d161087611 > > Another issue I see with this, I tell ruby configure about it, but > there is a MSVCRT chooser script that can fake/hide the trick for us. > > I'm starting to think that is we are letting MSVCR90 be the default > runtime, why I should bother with MinGW at all. > > I should be building Ruby with VC2008 instead... That seems like quite a lot of work for the linux cross compiling crowd. Of course, Charlie would probably say they don't need to link against msvcr90 they can just link against msvcrt6 and it should work :) My concern with it is would we have to be distributing tons of diferent msvcrt runtimes [like v 7.1 for the 2005 folks, etc.], if they link against those? Is there any benefit with trying to not use msvcrt6 from VC6? I hear rumors that it isn't as thread safe? Is it slower do we know? I suppose the reality is that most ruby programmers [and extension developers] at on Linux or OS X, so perhaps we should cater to them by making msvcrt.dll 6 the default? I prefer mingw because of the speed, but that's just my personal opinion. The cross compile stuff is nice, too. I'm ok with sticking with msvcrt6. Thoughts? -=r From rogerdpack at gmail.com Mon Mar 16 07:45:13 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Mon, 16 Mar 2009 05:45:13 -0600 Subject: [Rubyinstaller-devel] question Message-ID: I suppose it's all right for me to redistribute some alpha releases of the mingw installer? Just in case people want bleeding edge :P Also this post: http://www.ruby-forum.com/topic/179237#787348 mentions that they couldn't daemonize their mongrels under windows. Wonder why. Thanks. -r From will at hotgazpacho.org Mon Mar 16 10:04:50 2009 From: will at hotgazpacho.org (William Green) Date: Mon, 16 Mar 2009 10:04:50 -0400 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: References: <49AF7F4C.1020809@savagexi.com> <49B84F36.9060301@savagexi.com> <49B880F2.9060203@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> Message-ID: <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> MSVC6 is no longer available from Microsoft (I've even checked my Enterprisey MSDN Account... it is simply not available). This severely limits the number of people who can build Ruby on Windows, if the dependency is kept. I'm pretty sure this is why Luis embarked on the mingw32 build in the first place. If we MUST be dependent on an MS-provided C library, why not make it one of the ones that they distribute freely (with an IDE): 2005 or 2008. Question is, though, is this truly a dependency? On Mon, Mar 16, 2009 at 7:40 AM, Roger Pack wrote: > On Sun, Mar 15, 2009 at 12:40 PM, Luis Lavena > wrote: > > On Sun, Mar 15, 2009 at 12:56 AM, Roger Pack > wrote: > >> Perhaps we should [regardless of whether naming changes are approved > >> or not] by default link against msvcr90 [from the mingw builds], and > >> also include the VC2008 runtimes. That blesses the lives of the > >> VC2008 people, since they don't have to worry quite as much about > >> differing runtime library incompats. And it probably wouldn't be too > >> hard to link against it from mingw. Is it possible using the cross > >> compiler tho? > >> Thoughts? > >> -=r > >> > > > > AFAIK, none of the mingw32 packages for the different linux > > distributions bundle msvcr90 libraries. > > > > Without those, after installing (macport or apt-get package) manually > > find/build/install the msvcr90 libraries to properly link to the > > specified runtime. > > > > After that, if the user wanted to execute using Wine (to run > > specs/test and verify the build doesn't segfault) they will need to > > follow the winetricks: > > > > http://wiki.winehq.org/FAQ#head-a0aa6667e9ba22fad996f7f0d51dc7d161087611 > > > > Another issue I see with this, I tell ruby configure about it, but > > there is a MSVCRT chooser script that can fake/hide the trick for us. > > > > I'm starting to think that is we are letting MSVCR90 be the default > > runtime, why I should bother with MinGW at all. > > > > I should be building Ruby with VC2008 instead... > > That seems like quite a lot of work for the linux cross compiling > crowd. Of course, Charlie would probably say they don't need to link > against msvcr90 they can just link against msvcrt6 and it should work > :) > > My concern with it is would we have to be distributing tons of > diferent msvcrt runtimes [like v 7.1 for the 2005 folks, etc.], if > they link against those? Is there any benefit with trying to not use > msvcrt6 from VC6? I hear rumors that it isn't as thread safe? Is it > slower do we know? > > I suppose the reality is that most ruby programmers [and extension > developers] at on Linux or OS X, so perhaps we should cater to them by > making msvcrt.dll 6 the default? > I prefer mingw because of the speed, but that's just my personal > opinion. The cross compile stuff is nice, too. I'm ok with sticking > with msvcrt6. > Thoughts? > -=r > _______________________________________________ > Rubyinstaller-devel mailing list > Rubyinstaller-devel at rubyforge.org > http://rubyforge.org/mailman/listinfo/rubyinstaller-devel > -- -- Will Green http://willgreen.mp/ -------------- next part -------------- An HTML attachment was scrubbed... URL: From hramrach at centrum.cz Mon Mar 16 11:44:16 2009 From: hramrach at centrum.cz (Michal Suchanek) Date: Mon, 16 Mar 2009 16:44:16 +0100 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <49B880F2.9060203@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> Message-ID: 2009/3/16 William Green : > MSVC6 is no longer available from Microsoft (I've even checked my > Enterprisey MSDN Account... it is simply not available). This severely > limits the number of people who can build Ruby on Windows, if the dependency > is kept. I'm pretty sure this is why Luis embarked on the mingw32 build in > the first place. > If we MUST be?dependent?on an MS-provided C library, why not make it one of > the ones that they distribute freely (with an IDE): 2005 or 2008. > Question is, though, is this truly a dependency? Yes, it is. These are not installed in Windows (nor Wine) by default and you have to distribute them with Ruby for it to work. And they are some special beast so just downloading the dll and dropping it into the ruby directory is not enough if you forget to bundle it. On the other hand, the VC6 runtime is in all Windows so it is not a dependency. Thanks Michal From will at hotgazpacho.org Mon Mar 16 12:15:20 2009 From: will at hotgazpacho.org (William Green) Date: Mon, 16 Mar 2009 12:15:20 -0400 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: References: <49AF7F4C.1020809@savagexi.com> <49B880F2.9060203@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> Message-ID: <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> Michael, VC6 runtime may be on "all Windows", or available for download from Microsoft, but the VC6 development tools (specifically, the compiler and linker) are not available any longer, and I highly doubt they would even run on Vista. So, unless you already have a copy of the VC6 dev tools, you cannot obtain one now. This, I believe, is the reason for the MingW32 build in the first place; all the build tools are available, and not controlled by Microsoft. Luis, am I totally off-base here, am I missing something? On Mon, Mar 16, 2009 at 11:44 AM, Michal Suchanek wrote: > 2009/3/16 William Green : > > MSVC6 is no longer available from Microsoft (I've even checked my > > Enterprisey MSDN Account... it is simply not available). This severely > > limits the number of people who can build Ruby on Windows, if the > dependency > > is kept. I'm pretty sure this is why Luis embarked on the mingw32 build > in > > the first place. > > If we MUST be dependent on an MS-provided C library, why not make it one > of > > the ones that they distribute freely (with an IDE): 2005 or 2008. > > Question is, though, is this truly a dependency? > > Yes, it is. These are not installed in Windows (nor Wine) by default > and you have to distribute them with Ruby for it to work. And they are > some special beast so just downloading the dll and dropping it into > the ruby directory is not enough if you forget to bundle it. > > On the other hand, the VC6 runtime is in all Windows so it is not a > dependency. > > Thanks > > Michal > _______________________________________________ > Rubyinstaller-devel mailing list > Rubyinstaller-devel at rubyforge.org > http://rubyforge.org/mailman/listinfo/rubyinstaller-devel > -- -- Will Green http://willgreen.mp/ -------------- next part -------------- An HTML attachment was scrubbed... URL: From hramrach at centrum.cz Mon Mar 16 13:53:47 2009 From: hramrach at centrum.cz (Michal Suchanek) Date: Mon, 16 Mar 2009 18:53:47 +0100 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> Message-ID: 2009/3/16 William Green : > Michael, > VC6 runtime may be on "all Windows", or available for download from > Microsoft, but the VC6 development tools (specifically, the compiler and > linker) are not available any longer, and I highly doubt they would even run > on Vista. So, unless you already have a copy of the VC6 dev tools, you > cannot obtain one now. > This, I believe, is the reason for the MingW32 build in the first place; all > the build tools are available, and not controlled by Microsoft. > Luis, am I totally off-base here, am I missing something? Yes, that's the reason for building with MinGW32. However, the compiler normally links with the VC6 runtime and there were suggestions to link with later runtime instead. This would make the process of building and distributing Ruby more troublesome in my view which is what the migw build process was supposed to avoid. Unlike the VC6 runtime which really is in all windows even the possibility to download the runtime from Microsoft is not of much use for the VC2008 runtime. You might be able to install the runtime system-wide if you are the administrator but just downloading it does not help you. Thanks Michal From luislavena at gmail.com Mon Mar 16 15:10:58 2009 From: luislavena at gmail.com (Luis Lavena) Date: Mon, 16 Mar 2009 15:10:58 -0400 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> Message-ID: <71166b3b0903161210v73c27186t1c00fdb1a82a55a7@mail.gmail.com> 2009/3/16 William Green : > Michael, > > VC6 runtime may be on "all Windows", or available for download from > Microsoft, but the VC6 development tools (specifically, the compiler and > linker) are not available any longer, and I highly doubt they would even run > on Vista. So, unless you already have a copy of the VC6 dev tools, you > cannot obtain one now. > To put this in better words: MSVCRT, Microsoft Visual C Runtime library is available in all the Windows versions (since 95 AFAIK). MSVCRT.dll, often defined as version 6.0 has been available since NT4 The real thing is that MSVCRT cannot be updated with new API since breaking that library will break all the programs that link to it. Anyhow, MS managed to make 7.0 version of that DLL in XP Service Pack 2 (7.0.2600.2180), so is not out dated as someone pointed before. The new development and new features cannot happen in that version, that's why MS built MSVCR80 and MSVCR90. > This, I believe, is the reason for the MingW32 build in the first place; all > the build tools are available, and not controlled by Microsoft. > Yes, MinGW is a freely available alternative that links to the same C Runtime library. Even more, the advantage has been proven building gems for Windows from Linux or OSX. And ignore for a second the speed stuff. > Luis, am I totally off-base here, am I missing something? However, the problem with that is not the compiler, but Ruby. Ruby uses "mingw32" and "mswin32" to identify the platform used to build the interpreter, so that defeats every attempt to interchange the compilers. Even more, it prepends the runtime used "msvcrt" to the dll of ruby: msvcrt-ruby18.dll for both mingw and mswin32, and "msvcr90" for ruby build with VC2008. I'm kind of getting bored by this whole discussion, since I already had this like 3 years ago, and the year after, and last year. It doesn't matter if we do MinGW/GCC or VC2008 or VC3000, changes will need to happen in both the development of Ruby and the gems being created by others, so, what will be the less painful process for everybody? -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From rogerdpack at gmail.com Wed Mar 18 09:54:09 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Wed, 18 Mar 2009 07:54:09 -0600 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <71166b3b0903161210v73c27186t1c00fdb1a82a55a7@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> <71166b3b0903161210v73c27186t1c00fdb1a82a55a7@mail.gmail.com> Message-ID: So perhaps somebody could answer some questions for me I'm a little confused. I hear things like "you can compile against any version of msvcrt.dll you want" and then...the question is you don't actually need to have that specific version of msvcrt.dll installed in the system, by default it will just "use" the version that's contained in msvcrt-ruby....dll Is that right? does "msvcrt-ruby...dll" contain a copy of msvcrt.dll within itself? Also, was there any concensus on whether errno differences were fatal? Sorry to bore Luis with this stuff :) And in the end I don't care what happens. I'd be happy to spearhead any devkit gem, as well. Thanks! -=r On Mon, Mar 16, 2009 at 1:10 PM, Luis Lavena wrote: > 2009/3/16 William Green : >> Michael, >> >> VC6 runtime may be on "all Windows", or available for download from >> Microsoft, but the VC6 development tools (specifically, the compiler and >> linker) are not available any longer, and I highly doubt they would even run >> on Vista. So, unless you already have a copy of the VC6 dev tools, you >> cannot obtain one now. >> > > To put this in better words: > > MSVCRT, Microsoft Visual C Runtime library is available in all the > Windows versions (since 95 AFAIK). > > MSVCRT.dll, often defined as version 6.0 has been available since NT4 > > The real thing is that MSVCRT cannot be updated with new API since > breaking that library will break all the programs that link to it. > > Anyhow, MS managed to make 7.0 version of that DLL in XP Service Pack > 2 (7.0.2600.2180), so is not out dated as someone pointed before. > > The new development and new features cannot happen in that version, > that's why MS built MSVCR80 and MSVCR90. > >> This, I believe, is the reason for the MingW32 build in the first place; all >> the build tools are available, and not controlled by Microsoft. >> > > Yes, MinGW is a freely available alternative that links to the same C > Runtime library. > > Even more, the advantage has been proven building gems for Windows > from Linux or OSX. > > And ignore for a second the speed stuff. > >> Luis, am I totally off-base here, am I missing something? > > However, the problem with that is not the compiler, but Ruby. Ruby > uses "mingw32" and "mswin32" to identify the platform used to build > the interpreter, so that defeats every attempt to interchange the > compilers. > > Even more, it prepends the runtime used "msvcrt" to the dll of ruby: > msvcrt-ruby18.dll for both mingw and mswin32, and "msvcr90" for ruby > build with VC2008. > > I'm kind of getting bored by this whole discussion, since I already > had this like 3 years ago, and the year after, and last year. > > It doesn't matter if we do MinGW/GCC or VC2008 or VC3000, changes will > need to happen in both the development of Ruby and the gems being > created by others, so, what will be the less painful process for > everybody? > -- > Luis Lavena > AREA 17 > - > Perfection in design is achieved not when there is nothing more to add, > but rather when there is nothing more to take away. > Antoine de Saint-Exup?ry > _______________________________________________ > Rubyinstaller-devel mailing list > Rubyinstaller-devel at rubyforge.org > http://rubyforge.org/mailman/listinfo/rubyinstaller-devel > From now at bitwi.se Wed Mar 18 11:04:05 2009 From: now at bitwi.se (Nikolai Weibull) Date: Wed, 18 Mar 2009 16:04:05 +0100 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby Message-ID: I just downloaded, built, and installed the one-click installer and it seems that require 'zlib' (as done by, for example, RubyGems) causes Ruby to look up. I see no reason for this behavior. Copying over zlib.so from the old installer seems to work fine. Also, rubygems_mingw/text.rb doesn't get included in the installer. From luislavena at gmail.com Wed Mar 18 12:26:08 2009 From: luislavena at gmail.com (Luis Lavena) Date: Wed, 18 Mar 2009 13:26:08 -0300 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: References: Message-ID: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> On Wed, Mar 18, 2009 at 12:04 PM, Nikolai Weibull wrote: > I just downloaded, built, and installed the one-click installer and it > seems that > This one, right? http://github.com/oneclick/rubyinstaller > require 'zlib' > > (as done by, for example, RubyGems) causes Ruby to look up. ?I see no > reason for this behavior. ?Copying over zlib.so from the old installer > seems to work fine. I'll investigate what happens on this. Perhaps an older version of zlib1.dll is interfering. > Also, rubygems_mingw/text.rb doesn't get included in the installer. rubygems has no text.rb file I'm aware of. And for the MSI installer contain those files, the WiX XML files need to be updated. -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From now at bitwi.se Wed Mar 18 13:52:52 2009 From: now at bitwi.se (Nikolai Weibull) Date: Wed, 18 Mar 2009 18:52:52 +0100 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> References: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> Message-ID: On Wed, Mar 18, 2009 at 17:26, Luis Lavena wrote: > On Wed, Mar 18, 2009 at 12:04 PM, Nikolai Weibull wrote: >> I just downloaded, built, and installed the one-click installer and it >> seems that > This one, right? > > http://github.com/oneclick/rubyinstaller Yes. >> Also, rubygems_mingw/text.rb doesn't get included in the installer. > > rubygems has no text.rb file I'm aware of. It does now. I cloned the repository (sorry about using the word "downloaded" earlier), did % rake CHECKOUT=1 % rake package and then installed using the generated msi. I should note that in the RubyGems that this downloaded there was a problem with the --destdir command line option to setup.rb, but the RubyGems developers seem to be aware of the problem and are hopefully going to fix it. Perhaps I shouldn't have done CHECKOUT=1? From rogerdpack at gmail.com Thu Mar 19 14:43:23 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Thu, 19 Mar 2009 12:43:23 -0600 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: References: Message-ID: do you have other versions of zlib*.dll on your system perhaps? -=r On Wed, Mar 18, 2009 at 9:04 AM, Nikolai Weibull wrote: > I just downloaded, built, and installed the one-click installer and it > seems that > > require 'zlib' > > (as done by, for example, RubyGems) causes Ruby to look up. ?I see no > reason for this behavior. ?Copying over zlib.so from the old installer > seems to work fine. > > Also, rubygems_mingw/text.rb doesn't get included in the installer. > _______________________________________________ > Rubyinstaller-devel mailing list > Rubyinstaller-devel at rubyforge.org > http://rubyforge.org/mailman/listinfo/rubyinstaller-devel > From luislavena at gmail.com Thu Mar 19 14:58:21 2009 From: luislavena at gmail.com (Luis Lavena) Date: Thu, 19 Mar 2009 15:58:21 -0300 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: References: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> Message-ID: <71166b3b0903191158s8fb1540w97d75a6f9b13008f@mail.gmail.com> On Wed, Mar 18, 2009 at 2:52 PM, Nikolai Weibull wrote: > On Wed, Mar 18, 2009 at 17:26, Luis Lavena wrote: > >> On Wed, Mar 18, 2009 at 12:04 PM, Nikolai Weibull wrote: > >>> I just downloaded, built, and installed the one-click installer and it >>> seems that > >> This one, right? >> >> http://github.com/oneclick/rubyinstaller > > Yes. > Cool. >>> Also, rubygems_mingw/text.rb doesn't get included in the installer. >> >> rubygems has no text.rb file I'm aware of. > > It does now. ?I cloned the repository (sorry about using the word > "downloaded" earlier), did > No, there is no text.rb file in RubyGems repository: http://rubygems.rubyforge.org/svn/trunk/ > % rake CHECKOUT=1 > % rake package > > and then installed using the generated msi. ?I should note that in the > RubyGems that this downloaded there was a problem with the --destdir > command line option to setup.rb, but the RubyGems developers seem to > be aware of the problem and are hopefully going to fix it. > > Perhaps I shouldn't have done CHECKOUT=1? There are a lot of moving parts that need resync, RubyGems, Ruby and others needs to be reviewed. Will have time over this weekend to resync master. I recommend not building the Microsoft Installer (MSI) package, since the files changed are not being included in the manifest. Just copy rubygems_mingw over ruby_mingw and you have a ruby version to relocate to any path you want. -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From now at bitwi.se Thu Mar 19 15:54:26 2009 From: now at bitwi.se (Nikolai Weibull) Date: Thu, 19 Mar 2009 20:54:26 +0100 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: <71166b3b0903191158s8fb1540w97d75a6f9b13008f@mail.gmail.com> References: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> <71166b3b0903191158s8fb1540w97d75a6f9b13008f@mail.gmail.com> Message-ID: On Thu, Mar 19, 2009 at 19:58, Luis Lavena wrote: > No, there is no text.rb file in RubyGems repository: > > http://rubygems.rubyforge.org/svn/trunk/ Well it certainly did here. Perhaps a bogus rubygems version got pulled for some reason. As I said, there were a lot of issues with the rubygems that I got by the commands below: >> % rake CHECKOUT=1 >> % rake package > There are a lot of moving parts that need resync, RubyGems, Ruby and > others needs to be reviewed. > Just copy rubygems_mingw over ruby_mingw and you have a ruby version > to relocate to any path you want. That's what I wound up doing :-) Anyway, I've gotten most things set up and am looking to get a win32 release of character-encodings out the door soon. The new ruby installer really looks promising. So far I have the feeling that the mingw-based build is one or two times faster than the latest ruby installer. Thanks! From luislavena at gmail.com Thu Mar 19 17:06:47 2009 From: luislavena at gmail.com (Luis Lavena) Date: Thu, 19 Mar 2009 18:06:47 -0300 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: References: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> <71166b3b0903191158s8fb1540w97d75a6f9b13008f@mail.gmail.com> Message-ID: <71166b3b0903191406t5794a728v38559a20fb7554af@mail.gmail.com> On Thu, Mar 19, 2009 at 4:54 PM, Nikolai Weibull wrote: > On Thu, Mar 19, 2009 at 19:58, Luis Lavena wrote: > >> No, there is no text.rb file in RubyGems repository: >> >> http://rubygems.rubyforge.org/svn/trunk/ > > Well it certainly did here. ?Perhaps a bogus rubygems version got > pulled for some reason. ?As I said, there were a lot of issues with > the rubygems that I got by the commands below: > >>> % rake CHECKOUT=1 >>> % rake package > Definitely I'll take a look to that, sorry for the problems. >> There are a lot of moving parts that need resync, RubyGems, Ruby and >> others needs to be reviewed. > >> Just copy rubygems_mingw over ruby_mingw and you have a ruby version >> to relocate to any path you want. > > That's what I wound up doing :-) > :-P > Anyway, I've gotten most things set up and am looking to get a win32 > release of character-encodings out the door soon. ?The new ruby > installer really looks promising. ?So far I have the feeling that the > mingw-based build is one or two times faster than the latest ruby > installer. ?Thanks! character-encodings you said? great! I hear some guys looking for that gem! :-D Had that in my todo list for the past weeks, but job took over my free time (and I was traveling a lot). Keep us posted and let us know anything that we can do to help you on this. Cheers! -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From now at bitwi.se Thu Mar 19 17:26:59 2009 From: now at bitwi.se (Nikolai Weibull) Date: Thu, 19 Mar 2009 22:26:59 +0100 Subject: [Rubyinstaller-devel] Current Zlib module hangs Ruby In-Reply-To: <71166b3b0903191406t5794a728v38559a20fb7554af@mail.gmail.com> References: <71166b3b0903180926l316bcb9ftcb6a99267e45d82e@mail.gmail.com> <71166b3b0903191158s8fb1540w97d75a6f9b13008f@mail.gmail.com> <71166b3b0903191406t5794a728v38559a20fb7554af@mail.gmail.com> Message-ID: On Thu, Mar 19, 2009 at 22:06, Luis Lavena wrote: > On Thu, Mar 19, 2009 at 4:54 PM, Nikolai Weibull wrote: >> Well it certainly did here. ?Perhaps a bogus rubygems version got >> pulled for some reason. ?As I said, there were a lot of issues with >> the rubygems that I got by the commands below: > Definitely I'll take a look to that, sorry for the problems. I'm guessing that the rubygems devs are to blame. >> Anyway, I've gotten most things set up and am looking to get a win32 >> release of character-encodings out the door soon. > character-encodings you said? great! > > I hear some guys looking for that gem! :-D > > Had that in my todo list for the past weeks, but job took over my free > time (and I was traveling a lot). Yeah, he's been mailing me for a while now. I caught the flu and also had to catch up with ?real? work. > Keep us posted and let us know anything that we can do to help you on this. Thanks; will do. From luislavena at gmail.com Sat Mar 21 12:15:06 2009 From: luislavena at gmail.com (Luis Lavena) Date: Sat, 21 Mar 2009 13:15:06 -0300 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: References: <49AF7F4C.1020809@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> <71166b3b0903161210v73c27186t1c00fdb1a82a55a7@mail.gmail.com> Message-ID: <71166b3b0903210915u2fe04894j99378e85630166da@mail.gmail.com> On Wed, Mar 18, 2009 at 10:54 AM, Roger Pack wrote: > > So perhaps somebody could answer some questions for me I'm a little confused. > I hear things like "you can compile against any version of msvcrt.dll you want" > and then...the question is > you don't actually need to have that specific version of msvcrt.dll > installed in the system, by default it will just "use" the version > that's contained in msvcrt-ruby....dll > Is that right? > does "msvcrt-ruby...dll" contain a copy of msvcrt.dll within itself? > msvcrt-ruby18.dll or msvcrt-ruby19.dll are not statically linked to MSVCRT, they are dynamically. Is not the linking what is the problem, but the proper memory management and the convention used by the share ruby library. > Also, was there any concensus on whether errno differences were fatal? I didn't remember about errno until Ruby-core guys pointed that again, bad memory from me, once more. > Sorry to bore Luis with this stuff :) Sorry to take me so long to get back to you, overwhelmed by work for the past weeks, trying to catch up with everything. > And in the end I don't care what happens. He, well, this whole stuff is to complicated and tiring. > I'd be happy to spearhead any devkit gem, as well. devkit gem will a wrapper around the mssy/gcc rake tasks of rubyinstaller compiler, with the only difference that they will set the "sandbox" to a specific folder and put some batch wrapper in your ruby/bin folder. > Thanks! > -=r > Cheers, -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From rogerdpack at gmail.com Sat Mar 21 12:50:54 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Sat, 21 Mar 2009 10:50:54 -0600 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <71166b3b0903210915u2fe04894j99378e85630166da@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> <71166b3b0903161210v73c27186t1c00fdb1a82a55a7@mail.gmail.com> <71166b3b0903210915u2fe04894j99378e85630166da@mail.gmail.com> Message-ID: On Sat, Mar 21, 2009 at 10:15 AM, Luis Lavena wrote: > On Wed, Mar 18, 2009 at 10:54 AM, Roger Pack wrote: >> >> So perhaps somebody could answer some questions for me I'm a little confused. >> I hear things like "you can compile against any version of msvcrt.dll you want" >> and then...the question is >> you don't actually need to have that specific version of msvcrt.dll >> installed in the system, by default it will just "use" the version >> that's contained in msvcrt-ruby....dll >> Is that right? >> does "msvcrt-ruby...dll" contain a copy of msvcrt.dll within itself? >> > > msvcrt-ruby18.dll or msvcrt-ruby19.dll are not statically linked to > MSVCRT, they are dynamically. > > Is not the linking what is the problem, but the proper memory > management and the convention used by the share ruby library. Ok so what I'm hearing is that say you install mingw ruby then you install gem install mysql --platform=mswin32 which we'll pretend was built using VC2008 it will by "use" msvcrt 6 dll for it [hence it needs to be careful with frees and such]? There is no need for the system to have msvcr9.dll on it at all? >> I'd be happy to spearhead any devkit gem, as well. > > devkit gem will a wrapper around the mssy/gcc rake tasks of > rubyinstaller compiler, with the only difference that they will set > the "sandbox" to a specific folder and put some batch wrapper in your > ruby/bin folder. how can I help? re: errno so does this mean that it basically *is* impossible to compile against other msvcrt.dll's? I.e. only one can be "blessed" as a time? thanks! -=r From rogerdpack at gmail.com Sat Mar 28 17:47:17 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Sat, 28 Mar 2009 15:47:17 -0600 Subject: [Rubyinstaller-devel] Moving Ruby 1.9.1 forward on Windows In-Reply-To: <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> References: <49AF7F4C.1020809@savagexi.com> <49BC09D8.8010604@savagexi.com> <71166b3b0903151140jafc5189r73ec74681d49de82@mail.gmail.com> <398d3e0e0903160704t139350f4t7d907f4cb5fff3af@mail.gmail.com> <398d3e0e0903160915x587ba982pc852db1307a18ff5@mail.gmail.com> Message-ID: So if my theory is correct [I don't have all the information, but here's my guess] 1) if you are "careful" then you code, even though it was compiled against a different version of msvcrt.dll, will run in *any* windows ruby environment. I assume being careful means that you avoid the use off errno [and calling anything that uses errno]. Does anyone know which versions actually differ in errno numbering? If that's the case then we should be able to release the mingw version and then windows extension developers use VC 2008 and then package up their stuff "and name it mingw" [similar to currently doing a gem install --platform=mswin32]. Am I off base? Thanks. -=r 2009/3/16 William Green : > Michael, > VC6 runtime may be on "all Windows", or available for download from > Microsoft, but the VC6 development tools (specifically, the compiler and > linker) are not available any longer, and I highly doubt they would even run > on Vista. So, unless you already have a copy of the VC6 dev tools, you > cannot obtain one now. > This, I believe, is the reason for the MingW32 build in the first place; all > the build tools are available, and not controlled by Microsoft. > Luis, am I totally off-base here, am I missing something? > > On Mon, Mar 16, 2009 at 11:44 AM, Michal Suchanek > wrote: >> >> 2009/3/16 William Green : >> > MSVC6 is no longer available from Microsoft (I've even checked my >> > Enterprisey MSDN Account... it is simply not available). This severely >> > limits the number of people who can build Ruby on Windows, if the >> > dependency >> > is kept. I'm pretty sure this is why Luis embarked on the mingw32 build >> > in >> > the first place. >> > If we MUST be?dependent?on an MS-provided C library, why not make it one >> > of >> > the ones that they distribute freely (with an IDE): 2005 or 2008. >> > Question is, though, is this truly a dependency? >> >> Yes, it is. These are not installed in Windows (nor Wine) by default >> and you have to distribute them with Ruby for it to work. And they are >> some special beast so just downloading the dll and dropping it into >> the ruby directory is not enough if you forget to bundle it. >> >> On the other hand, the VC6 runtime is in all Windows so it is not a >> dependency. >> >> Thanks >> >> Michal >> _______________________________________________ >> Rubyinstaller-devel mailing list >> Rubyinstaller-devel at rubyforge.org >> http://rubyforge.org/mailman/listinfo/rubyinstaller-devel > > > -- > -- > Will Green > http://willgreen.mp/ > > _______________________________________________ > Rubyinstaller-devel mailing list > Rubyinstaller-devel at rubyforge.org > http://rubyforge.org/mailman/listinfo/rubyinstaller-devel > From rogerdpack at gmail.com Mon Mar 30 09:43:13 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Mon, 30 Mar 2009 07:43:13 -0600 Subject: [Rubyinstaller-devel] GCC 4.3 Message-ID: If anyone's interested my fork has an option GCC43=1 to use GCC 43 instead of the default. http://github.com/rogerdpack/rubyinstaller/tree/master feel free to use or merge if useful. I originally tried also adding extra CFLAGS like "-march=native -O3" but the subsequent build broke [not sure which one caused it to do so]. Just building with GCC43=1 seems to work all right though. Cheers! -=r From deepjungle.maca at gmail.com Mon Mar 30 13:06:33 2009 From: deepjungle.maca at gmail.com (deepj) Date: Mon, 30 Mar 2009 19:06:33 +0200 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: References: Message-ID: <49D0FC19.8020500@gmail.com> I use Ruby 1.9.1 with minwg gcc 4.3.x a long time. It seems to be OK, only Eventmachine doesn't work fine :(. My mingw32 is from this place: http://sourceforge.net/projects/tdm-gcc/ Dne 30.3.2009 15:43, Roger Pack napsal(a): > If anyone's interested my fork has an option GCC43=1 to use GCC 43 > instead of the default. > http://github.com/rogerdpack/rubyinstaller/tree/master > feel free to use or merge if useful. > > I originally tried also adding extra CFLAGS like "-march=native -O3" > but the subsequent build broke [not sure which one caused it to do > so]. Just building with GCC43=1 seems to work all right though. > > Cheers! > -=r > _______________________________________________ > Rubyinstaller-devel mailing list > Rubyinstaller-devel at rubyforge.org > http://rubyforge.org/mailman/listinfo/rubyinstaller-devel From rogerdpack at gmail.com Tue Mar 31 11:50:51 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Tue, 31 Mar 2009 09:50:51 -0600 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: <49D0FC19.8020500@gmail.com> References: <49D0FC19.8020500@gmail.com> Message-ID: On Mon, Mar 30, 2009 at 11:06 AM, deepj wrote: > I use Ruby 1.9.1 with minwg gcc 4.3.x a long time. It seems to be OK, only > Eventmachine doesn't work fine :(. > > My mingw32 is from this place: > http://sourceforge.net/projects/tdm-gcc/ Yeah that's the one I linked to as an option :) How to reproduce the EM errror? -=r From rogerdpack at gmail.com Tue Mar 31 12:37:22 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Tue, 31 Mar 2009 10:37:22 -0600 Subject: [Rubyinstaller-devel] .so versus .dll files Message-ID: Question for anybody. So...currently in ruby on windows it seems that it handles .so files quite nicely [in 1.9, .dll is actually no longer handled--only .so] Typical windows programs use .dll's, while Ruby uses .so's. How is this possible? Is it special? Thanks! -=r From luislavena at gmail.com Tue Mar 31 13:08:04 2009 From: luislavena at gmail.com (Luis Lavena) Date: Tue, 31 Mar 2009 14:08:04 -0300 Subject: [Rubyinstaller-devel] .so versus .dll files In-Reply-To: References: Message-ID: <71166b3b0903311008ue1e9166qc085bf11d8d5b074@mail.gmail.com> On Tue, Mar 31, 2009 at 1:37 PM, Roger Pack wrote: > Question for anybody. > So...currently in ruby on windows it seems that it handles .so files > quite nicely [in 1.9, .dll is actually no longer handled--only .so] > Typical windows programs use .dll's, while Ruby uses .so's. ?How is > this possible? ?Is it special? Actually is just the DLL name. they used to put that since there are extensions like zlib that shared the exact same name with the DLL (zlib.dll back in 1.x times, now is zlib1.dll). There are others, like readline and maybe others (that I don't remember right now). So looking just for .so is like Python looking for .pyd files instead of plain DLL ones. Inside of these files, there is just one exported function "Init_[extension_name]" > Thanks! Cheers, -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From luislavena at gmail.com Tue Mar 31 14:09:51 2009 From: luislavena at gmail.com (Luis Lavena) Date: Tue, 31 Mar 2009 15:09:51 -0300 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: References: Message-ID: <71166b3b0903311109n488552a3sd7a309058192f213@mail.gmail.com> On Mon, Mar 30, 2009 at 10:43 AM, Roger Pack wrote: > > If anyone's interested my fork has an option GCC43=1 to use GCC 43 > instead of the default. > http://github.com/rogerdpack/rubyinstaller/tree/master > feel free to use or merge if useful. > Awesome work Roger!. Definitely I'm going to merge those into dsl branch. > I originally tried also adding extra CFLAGS like "-march=native -O3" > but the subsequent build broke [not sure which one caused it to do > so]. ?Just building with GCC43=1 seems to work all right though. > Yeah, native optimizations are not 100% working, tried that once and reverted to -O2. I'm going to work over the next few days on testing GCC4 and GCC 3.4.5 builds of Ruby. Trying to get mingw-w64 working too, had a Xeon at the office for a couple of days and looks like my migration to x64 should happen soon. -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From rogerdpack at gmail.com Tue Mar 31 15:45:31 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Tue, 31 Mar 2009 13:45:31 -0600 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: <71166b3b0903311109n488552a3sd7a309058192f213@mail.gmail.com> References: <71166b3b0903311109n488552a3sd7a309058192f213@mail.gmail.com> Message-ID: > Yeah, native optimizations are not 100% working, tried that once and > reverted to -O2. > > I'm going to work over the next few days on testing GCC4 and GCC 3.4.5 > builds of Ruby. > > Trying to get mingw-w64 working too, had a Xeon at the office for a > couple of days and looks like my migration to x64 should happen soon. Wow good luck with that one. :) -=r From deepjungle.maca at gmail.com Tue Mar 31 16:47:28 2009 From: deepjungle.maca at gmail.com (deepj) Date: Tue, 31 Mar 2009 22:47:28 +0200 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: References: <49D0FC19.8020500@gmail.com> Message-ID: <49D28160.7010008@gmail.com> > Yeah that's the one I linked to as an option :) > How to reproduce the EM errror? You can run EventMachine tests using rake:test. When I run it EventMachine will stop responding (I must kill it). The compile seems to be fine, only a few warnings. Maybe the problem is in OpenSSL libraries. I've used OpenSSL 0.9.8j from this place http://www.slproweb.com/products/Win32OpenSSL.html (the full installation contains dev files for mingw) when I compiled Ruby 1.9.1. Can you try it and you can confirm it if the problem in this? I'm sorry for my ugly english :) From luislavena at gmail.com Tue Mar 31 17:52:13 2009 From: luislavena at gmail.com (Luis Lavena) Date: Tue, 31 Mar 2009 18:52:13 -0300 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: <49D28160.7010008@gmail.com> References: <49D0FC19.8020500@gmail.com> <49D28160.7010008@gmail.com> Message-ID: <71166b3b0903311452y272aaee4l2931231d95affc38@mail.gmail.com> On Tue, Mar 31, 2009 at 5:47 PM, deepj wrote: >> Yeah that's the one I linked to as an option :) >> How to reproduce the EM errror? > > You can run EventMachine tests using rake:test. When I run it EventMachine > will stop responding (I must kill it). The compile seems to be fine, only a > few warnings. > > Maybe the ?problem is in OpenSSL libraries. I've used OpenSSL 0.9.8j from > this place http://www.slproweb.com/products/Win32OpenSSL.html (the full > installation contains dev files for mingw) when I compiled Ruby 1.9.1. Can > you try it and you can confirm it if the problem in this? > > I'm sorry for my ugly english :) I believe that version of OpenSSL is linked with MSVCR80/90, not msvcrt, which is used by MinGW). Be careful using those binaries. -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry From rogerdpack at gmail.com Tue Mar 31 17:54:06 2009 From: rogerdpack at gmail.com (Roger Pack) Date: Tue, 31 Mar 2009 15:54:06 -0600 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: <49D28160.7010008@gmail.com> References: <49D0FC19.8020500@gmail.com> <49D28160.7010008@gmail.com> Message-ID: > Maybe the ?problem is in OpenSSL libraries. I've used OpenSSL 0.9.8j from > this place http://www.slproweb.com/products/Win32OpenSSL.html (the full > installation contains dev files for mingw) when I compiled Ruby 1.9.1. Can > you try it and you can confirm it if the problem in this? > > I'm sorry for my ugly english :) Sure I'll try it out sometime. Might take me a little bit to get to it :) -=r From luislavena at gmail.com Tue Mar 31 18:15:44 2009 From: luislavena at gmail.com (Luis Lavena) Date: Tue, 31 Mar 2009 19:15:44 -0300 Subject: [Rubyinstaller-devel] GCC 4.3 In-Reply-To: References: <49D0FC19.8020500@gmail.com> <49D28160.7010008@gmail.com> Message-ID: <71166b3b0903311515m50b1a36ep6111cb09fa61dec1@mail.gmail.com> On Tue, Mar 31, 2009 at 6:54 PM, Roger Pack wrote: >> Maybe the ?problem is in OpenSSL libraries. I've used OpenSSL 0.9.8j from >> this place http://www.slproweb.com/products/Win32OpenSSL.html (the full >> installation contains dev files for mingw) when I compiled Ruby 1.9.1. Can >> you try it and you can confirm it if the problem in this? >> >> I'm sorry for my ugly english :) > > > Sure I'll try it out sometime. ?Might take me a little bit to get to it :) What about proper-openssl branch? It builds it from source and do not depend on external binaries (so it can build latest version of OpenSSL). -- Luis Lavena AREA 17 - Perfection in design is achieved not when there is nothing more to add, but rather when there is nothing more to take away. Antoine de Saint-Exup?ry