Bug#309196: Bug#307231: RageIRCd build failure on s390
Package: rageircd Version: 2.0.0-2 Hi Marc, Apologies for not getting back to you sooner - I've been working hard. I've got around to the proper bug fix so bare with me. Firstly, replace all the current VA_COPY declaration statements in include/common.h with: #ifdef HAVE___VA_COPY #define VA_COPY __va_copy #elif defined(HAVE_VA_COPY) #define VA_COPY va_copy #else #define VA_COPY(x, y) x = y #endif Secondly, in configure.ac line 281: AC_CHECK_FUNCS([times index rindex setlinebuf]) Change to: AC_CHECK_FUNCS([times index rindex setlinebuf va_copy __va_copy]) Thirdly, re-generate the rage2 autoconf. For this just execute ./ config -a. You will see: persephone$ ./config -a Rebuilding configure... done! Rebuilding include/setup.h.in... done! From here, do a make distclean to erase any junk and re-configure as normal. Configure should detect the type of va_copy implementation on the host architecture and set itself up accordingly. :) Once again apologies for the delay. I know this e-mail isn't very professional of me but I haven't got time to back-port the changes to a clean 2.0.0 and patch. This code will be going into the BLUEMOON branch today. (Build 100 I think.) Regards Alasdair On 15 May 2005, at 15:20, Marc Haber wrote: Thiemo Seufer suggests replacing the entire VA_COPY detection code with: #if defined(__GNUC__) || (defined(__powerpc__) defined (__FreeBSD__)) #define VA_COPY __va_copy #elif defined(__powerpc__) defined(__NetBSD__) #define VA_COPY va_copy #else #define VA_COPY(x, y) x = y #endif With that page, a binary is created on s390 (cannot verify whether it is actually runnable), and a working binary is created on i386. Please indicate whether that fix would be appropriate for you as I don't want to have a package in Debian that deviates too far from your work. Greetings Marc -- -- --- Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#309196: Fwd: Bug#307231: RageIRCd build failure on s390
Package: rageircdVersion: 2.0.0-2Re. the below message, the configure.ac line number may be incorrect as autoconf has changed a bit between 2.0.0 and the current BLUEMOON revision. It shouldn't be too far out though.AlasdairBegin forwarded message:From: Alasdair McWilliam [EMAIL PROTECTED]Date: 16 May 2005 14:43:06 BDTTo: Marc Haber [EMAIL PROTECTED]Cc: [EMAIL PROTECTED]Subject: Re: Bug#307231: RageIRCd build failure on s390 Package: rageircdVersion: 2.0.0-2Hi Marc,Apologies for not getting back to you sooner - I've been working hard.I've got around to the proper bug fix so bare with me.Firstly, replace all the current VA_COPY declaration statements in include/common.h with:#ifdef HAVE___VA_COPY#define VA_COPY __va_copy#elif defined(HAVE_VA_COPY)#define VA_COPY va_copy#else#define VA_COPY(x, y) x = y#endifSecondly, in configure.ac line 281:AC_CHECK_FUNCS([times index rindex setlinebuf])Change to:AC_CHECK_FUNCS([times index rindex setlinebuf va_copy __va_copy])Thirdly, re-generate the rage2 autoconf. For this just execute ./config -a. You will see:persephone$ ./config -aRebuilding configure... done!Rebuilding include/setup.h.in... done!From here, do a make distclean to erase any junk and re-configure as normal. Configure should detect the type of va_copy implementation on the host architecture and set itself up accordingly. :)Once again apologies for the delay. I know this e-mail isn't very professional of me but I haven't got time to back-port the changes to a clean 2.0.0 and patch. This code will be going into the BLUEMOON branch today. (Build 100 I think.)RegardsAlasdairOn 15 May 2005, at 15:20, Marc Haber wrote: Thiemo Seufer suggests replacing the entire VA_COPY detection code with:#if defined(__GNUC__) || (defined(__powerpc__) defined(__FreeBSD__))#define VA_COPY __va_copy#elif defined(__powerpc__) defined(__NetBSD__)#define VA_COPY va_copy#else#define VA_COPY(x, y) x = y#endifWith that page, a binary is created on s390 (cannot verify whether itis actually runnable), and a working binary is created on i386.Please indicate whether that fix would be appropriate for you as Idon't want to have a package in Debian that deviates too far from yourwork.GreetingsMarc-- -Marc Haber | "I don't trust Computers. They | Mailadresse im HeaderMannheim, Germany | lose things." Winona Ryder | Fon: *49 621 72739834Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835
Bug#307231: Fwd: Bug#307231: RageIRCd build failure on s390
See messages forwarded below. Sent to the wrong bug report in error.Begin forwarded message:From: Alasdair McWilliam [EMAIL PROTECTED]Date: 16 May 2005 14:45:15 BDTTo: Marc Haber [EMAIL PROTECTED]Cc: [EMAIL PROTECTED]Subject: Fwd: Bug#307231: RageIRCd build failure on s390 Package: rageircdVersion: 2.0.0-2Re. the below message, the configure.ac line number may be incorrect as autoconf has changed a bit between 2.0.0 and the current BLUEMOON revision. It shouldn't be too far out though.AlasdairBegin forwarded message:From: Alasdair McWilliam [EMAIL PROTECTED]Date: 16 May 2005 14:43:06 BDTTo: Marc Haber [EMAIL PROTECTED]Cc: [EMAIL PROTECTED]Subject: Re: Bug#307231: RageIRCd build failure on s390 Package: rageircdVersion: 2.0.0-2Hi Marc,Apologies for not getting back to you sooner - I've been working hard.I've got around to the proper bug fix so bare with me.Firstly, replace all the current VA_COPY declaration statements in include/common.h with:#ifdef HAVE___VA_COPY#define VA_COPY __va_copy#elif defined(HAVE_VA_COPY)#define VA_COPY va_copy#else#define VA_COPY(x, y) x = y#endifSecondly, in configure.ac line 281:AC_CHECK_FUNCS([times index rindex setlinebuf])Change to:AC_CHECK_FUNCS([times index rindex setlinebuf va_copy __va_copy])Thirdly, re-generate the rage2 autoconf. For this just execute ./config -a. You will see:persephone$ ./config -aRebuilding configure... done!Rebuilding include/setup.h.in... done!From here, do a make distclean to erase any junk and re-configure as normal. Configure should detect the type of va_copy implementation on the host architecture and set itself up accordingly. :)Once again apologies for the delay. I know this e-mail isn't very professional of me but I haven't got time to back-port the changes to a clean 2.0.0 and patch. This code will be going into the BLUEMOON branch today. (Build 100 I think.)RegardsAlasdairOn 15 May 2005, at 15:20, Marc Haber wrote: Thiemo Seufer suggests replacing the entire VA_COPY detection code with:#if defined(__GNUC__) || (defined(__powerpc__) defined(__FreeBSD__))#define VA_COPY __va_copy#elif defined(__powerpc__) defined(__NetBSD__)#define VA_COPY va_copy#else#define VA_COPY(x, y) x = y#endifWith that page, a binary is created on s390 (cannot verify whether itis actually runnable), and a working binary is created on i386.Please indicate whether that fix would be appropriate for you as Idon't want to have a package in Debian that deviates too far from yourwork.GreetingsMarc-- -Marc Haber | "I don't trust Computers. They | Mailadresse im HeaderMannheim, Germany | lose things." Winona Ryder | Fon: *49 621 72739834Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835
Bug#309196: Bug#307231: RageIRCd build failure on s390
* Alasdair McWilliam: Change to: AC_CHECK_FUNCS([times index rindex setlinebuf va_copy __va_copy]) va_copy is a macro, so this check won't work reliably. On Debian, it's safe to use va_copy unconditionally because it's part of GCC's stdarg.h header. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#309196: Bug#307231: RageIRCd build failure on s390
Hi Alasdair, On Mon, May 16, 2005 at 04:21:15PM +0200, Florian Weimer wrote: * Alasdair McWilliam: Change to: AC_CHECK_FUNCS([times index rindex setlinebuf va_copy __va_copy]) va_copy is a macro, so this check won't work reliably. Florian usually knows what he is talking about, so I'd rather refrain from applying your patch. On Debian, it's safe to use va_copy unconditionally because it's part of GCC's stdarg.h header. Knowing that rageircd is not only meant for Debian, I'd suggest using the code from Thiemo, patch repeated: --- ./include/common.h 2004-12-07 03:04:34.0 + +++ /tmp/dpep-work.aVeWMw/rageircd-2.0.0/include/common.h 2005-05-15 14:07:40.192791162 + @@ -48,17 +48,10 @@ #define ASSERT(xx) #endif -/* VA_COPY - namely for x86_64 and PowerPC architectures. - * Copyright Rossi 'vejeta' Marcello ([EMAIL PROTECTED]) - */ -#ifdef __powerpc__ -#ifdef __NetBSD__ -#define VA_COPY va_copy -#elif defined(__FreeBSD__) || defined(__linux__) -#define VA_COPY __va_copy -#endif -#elif defined(__x86_64) +#if defined(__GNUC__) || (defined(__powerpc__) defined(__FreeBSD__)) #define VA_COPY __va_copy +#elif defined(__powerpc__) defined(__NetBSD__) +#define VA_COPY va_copy #else #define VA_COPY(x, y) x = y #endif Would that be ok with both of you? Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#309196: Bug#307231: RageIRCd build failure on s390
* Marc Haber: +#if defined(__GNUC__) || (defined(__powerpc__) defined(__FreeBSD__)) #define VA_COPY __va_copy +#elif defined(__powerpc__) defined(__NetBSD__) +#define VA_COPY va_copy #else #define VA_COPY(x, y) x = y #endif Would that be ok with both of you? __va_copy is provided by GCC's stdarg.h as well, so this patch should be fine. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#309196: Bug#307231: RageIRCd build failure on s390
I expect you know more than me about this so I'll go with the patch you've given. Thanks for your input on this issue too it's really appreciated. :-) I'll re-fix the BLUEMOON branch tomorrow evening probably. Is it possible to close the other bug associated with rageircd as it's not really a bug? Cheers On 16 May 2005, at 17:16, Florian Weimer wrote: * Marc Haber: +#if defined(__GNUC__) || (defined(__powerpc__) defined (__FreeBSD__)) #define VA_COPY __va_copy +#elif defined(__powerpc__) defined(__NetBSD__) +#define VA_COPY va_copy #else #define VA_COPY(x, y) x = y #endif Would that be ok with both of you? __va_copy is provided by GCC's stdarg.h as well, so this patch should be fine. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#309196: Bug#307231: RageIRCd build failure on s390
Hi, On Mon, May 16, 2005 at 10:19:08PM +0100, Alasdair McWilliam wrote: I expect you know more than me about this so I'll go with the patch you've given. Thanks. So I can upload the patched version to unstable, fixing the s390 issue? Is it possible to close the other bug associated with rageircd as it's not really a bug? It is an issue which makes the current rageircd unsuitable for release with a stable version. We have missed sarge anyway, so the package wouldn't migrate from unstable to testing anyway. That bug report doesn't really hurt and is a reminder that the zlib issue still needs to be sorted out. Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#309196: Bug#307231: RageIRCd build failure on s390
Thanks. So I can upload the patched version to unstable, fixing the s390 issue? Yes :-) It is an issue which makes the current rageircd unsuitable for release with a stable version. We have missed sarge anyway, so the package wouldn't migrate from unstable to testing anyway. That bug report doesn't really hurt and is a reminder that the zlib issue still needs to be sorted out. Aye. (Any way the package can be pushed through quickly? Nobody will notice! ;p Hehehehehe.) lol. Regards Alasdair -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#307231: RageIRCd build failure on s390
Thiemo Seufer suggests replacing the entire VA_COPY detection code with: #if defined(__GNUC__) || (defined(__powerpc__) defined(__FreeBSD__)) #define VA_COPY __va_copy #elif defined(__powerpc__) defined(__NetBSD__) #define VA_COPY va_copy #else #define VA_COPY(x, y) x = y #endif With that page, a binary is created on s390 (cannot verify whether it is actually runnable), and a working binary is created on i386. Please indicate whether that fix would be appropriate for you as I don't want to have a package in Debian that deviates too far from your work. Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#307231: RageIRCd build failure on s390
tags #307231 thanks On Sun, May 01, 2005 at 11:44:56PM +0100, Alasdair McWilliam wrote: The compile failure is in src/ircsprintf.c as follows. cc -g -Wall -O2 -export-dynamic -I../adns -I../include -I../zlib -c ircsprintf.c ircsprintf.c: In function `ircvsprintf': ircsprintf.c:29: error: incompatible types in assignment ircsprintf.c: In function `ircvsnprintf': ircsprintf.c:123: error: incompatible types in assignment make[2]: *** [ircsprintf.o] Error 1 Confirmed. As I don't have access to a s390 system I cannot confirm the below bug fix. However, I believe the VA_COPY macro is the problem. Line 60 of include/common.h is: #elif defined(__x86_64) I believe if this is changed to the following statement, it will compile fine on s390. #elif defined(__x86_64) || defined(__s390) No. The error message is the same. Anything I can do to help debugging? Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#307231: RageIRCd build failure on s390
tags #307231 thanks To help debugging you can configure with --enable-developer (either manually with ./configure or specified in config when it asks for additional arguments). This makes compile a bit more anally retentive and will complain at more things. Hehe. Alasdair On 7 May 2005, at 17:13, Marc Haber wrote: tags #307231 thanks On Sun, May 01, 2005 at 11:44:56PM +0100, Alasdair McWilliam wrote: The compile failure is in src/ircsprintf.c as follows. cc -g -Wall -O2 -export-dynamic -I../adns -I../include -I../zlib -c ircsprintf.c ircsprintf.c: In function `ircvsprintf': ircsprintf.c:29: error: incompatible types in assignment ircsprintf.c: In function `ircvsnprintf': ircsprintf.c:123: error: incompatible types in assignment make[2]: *** [ircsprintf.o] Error 1 Confirmed. As I don't have access to a s390 system I cannot confirm the below bug fix. However, I believe the VA_COPY macro is the problem. Line 60 of include/common.h is: #elif defined(__x86_64) I believe if this is changed to the following statement, it will compile fine on s390. #elif defined(__x86_64) || defined(__s390) No. The error message is the same. Anything I can do to help debugging? Greetings Marc -- -- --- Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#307231: RageIRCd build failure on s390
Probably because I've made a wee booboo. ;-) 2.0.1 will be along quicker than planned as I was up until 3am doing lots of tweaking and fixing here and there. Give me a few days. (I've a better fix for the va_copy issue too, will use autoconf.) Thanks for responding Alasdair On 2 May 2005, at 06:36, Marc Haber wrote: severity #307231 serious thanks Justification: hidden FTBFS Hi, On Sun, May 01, 2005 at 11:44:56PM +0100, Alasdair McWilliam wrote: I've checked the build logs for rageircd 2.0.0-2 in unstable and have noticed that the actual rageircd binary doesn't build on the s390 architecture. Hence I am reporting it here. Thanks for noticing that. I am trying to reproduce the issue, and will then see whether your fix will actually fix the issue. Actually, the build should _fail_ in that case and not continue to deliver a non-working package. Why doesn't the make process issue a non-zero exit code here? Greetings Marc -- -- --- Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#307231: RageIRCd build failure on s390
Package: rageircdVersion: 2.0.0-2HiI've checked the build logs for rageircd 2.0.0-2 in unstable and have noticed that the actual rageircd binary doesn't build on the s390 architecture. Hence I am reporting it here.The compile failure is in src/ircsprintf.c as follows.cc -g -Wall -O2 -export-dynamic -I../adns -I../include -I../zlib -c ircsprintf.circsprintf.c: In function `ircvsprintf':ircsprintf.c:29: error: incompatible types in assignmentircsprintf.c: In function `ircvsnprintf':ircsprintf.c:123: error: incompatible types in assignmentmake[2]: *** [ircsprintf.o] Error 1As I don't have access to a s390 system I cannot confirm the below bug fix. However, I believe the VA_COPY macro is the problem. Line 60 of include/common.h is:#elif defined(__x86_64)I believe if this is changed to the following statement, it will compile fine on s390.#elif defined(__x86_64) || defined(__s390)All other architectures appear to compile successfully.I'm mailing the fix rather than providing a new version of rageircd 2.0 as I don't think the problem really calls for a new version just yet. I can't see many people with an s390 running RageIRCd. However, the fix is already in rageircd 2.0.1-alpha. Perhaps the Debian source will be patched and compile tested? If so I would appreciate feedback accordingly.RegardsAlasdairRageIRCd
Bug#307231: RageIRCd build failure on s390
severity #307231 serious thanks Justification: hidden FTBFS Hi, On Sun, May 01, 2005 at 11:44:56PM +0100, Alasdair McWilliam wrote: I've checked the build logs for rageircd 2.0.0-2 in unstable and have noticed that the actual rageircd binary doesn't build on the s390 architecture. Hence I am reporting it here. Thanks for noticing that. I am trying to reproduce the issue, and will then see whether your fix will actually fix the issue. Actually, the build should _fail_ in that case and not continue to deliver a non-working package. Why doesn't the make process issue a non-zero exit code here? Greetings Marc -- - Marc Haber | I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things.Winona Ryder | Fon: *49 621 72739834 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 72739835 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]