Re: Can't build -CURRENT on 4.7
Ruslan Ermilov wrote: On Fri, Jun 06, 2003 at 11:57:00PM -0700, David O'Brien wrote: On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Committers are not required to support building 5-CURRENT, post 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the change. However, someone will probably patch the build system to tolerate it. Please correct me if I'm wrong, but I thought that this support will no longer be _required_ when we have a first release on the RELENG_5 (-STABLE) branch. [EMAIL PROTECTED] That was my understanding too, alas. Granted, we *must* provide an update path from the last 4.x release to the first *stable* 5.x release. But I do recall being advised that the core decided to support upgrade from *any* 4.x release to 5.0. -- Daniel C. Sobral (8-DCS) Gerencia de Operacoes Divisao de Comunicacao de Dados Coordenacao de Seguranca VIVO Centro Oeste Norte Fones: 55-61-313-7654/Cel: 55-61-9618-0904 E-mail: [EMAIL PROTECTED] [EMAIL PROTECTED] [EMAIL PROTECTED] Outros: [EMAIL PROTECTED] [EMAIL PROTECTED] [EMAIL PROTECTED] You can move the world with an idea, but you have to think of it first. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
On Sat, Jun 07, 2003 at 03:34:08PM -0700, Tim Kientzle wrote: David O'Brien wrote: This won't work on non-i386, due to alloca issues. + WORLDTMP=${WORLDTMP} CSTD= \ may. Hmmm... This seems like the Right Thing in any case, since it is one less assumption you're making about the build environment. I'm still getting buildworld failures, though. Long after the bootstrap, using the new tools, I'm seeing consistent failures in libpthread: building shared library libkse.so.1 thr_libc.So: In function `sigaction': thr_libc.So(.text+0x54): multiple definition of `_sigaction' thr_sigaction.So:/usr/src/current/lib/libpthread/thread/thr_sigaction.c:43: first defined here thr_libc.So: In function `sigprocmask': thr_libc.So(.text+0x34): multiple definition of `_sigprocmask' thr_sigprocmask.So:/usr/src/current/lib/libpthread/thread/thr_sigprocmask.c:46: first defined here *** Error code 1 Stop in /usr/src/current/lib/libpthread. *** Error code 1 Tim, These have been fixed just recently (see the latest committs to src/Makefile.inc1 and src/lib/libpthread), and the buildworld for 5.x on 4.x now works again (verified). There are some uncommitted patches floating around that allow for a buildkernel to survive as well. I have left them for Warner's consideration. Cheers, -- Ruslan Ermilov Sysadmin and DBA, [EMAIL PROTECTED] Sunbay Software Ltd, [EMAIL PROTECTED] FreeBSD committer pgp0.pgp Description: PGP signature
Re: Can't build -CURRENT on 4.7
On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Committers are not required to support building 5-CURRENT, post 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the change. However, someone will probably patch the build system to tolerate it. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
On Fri, Jun 06, 2003 at 11:57:00PM -0700, David O'Brien wrote: On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Committers are not required to support building 5-CURRENT, post 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the change. However, someone will probably patch the build system to tolerate it. That's a bit of an early call given that 5.x world is currently broken with the same error, as you know. Kris pgp0.pgp Description: PGP signature
Re: Can't build -CURRENT on 4.7
On Sat, Jun 07, 2003 at 12:08:31AM -0700, Kris Kennaway wrote: On Fri, Jun 06, 2003 at 11:57:00PM -0700, David O'Brien wrote: On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Committers are not required to support building 5-CURRENT, post 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the change. However, someone will probably patch the build system to tolerate it. That's a bit of an early call given that 5.x world is currently broken with the same error, as you know. I carefully worded the reply to specifically address build 5-CURRENT on 4.7. Can you try src/share/mk/bsd.sys.mk rev 1.29 to see if it fixes your problem? -- -- David ([EMAIL PROTECTED]) ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
On Sat, Jun 07, 2003 at 01:07:24AM -0700, David O'Brien wrote: I carefully worded the reply to specifically address build 5-CURRENT on 4.7. Can you try src/share/mk/bsd.sys.mk rev 1.29 to see if it fixes your problem? Have you tested it thoroughly? Didn't you back out -std=c99 in a previous revision because it also caused problems? Kris pgp0.pgp Description: PGP signature
Re: Can't build -CURRENT on 4.7
On Fri, Jun 06, 2003 at 11:57:00PM -0700, David O'Brien wrote: On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Committers are not required to support building 5-CURRENT, post 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the change. However, someone will probably patch the build system to tolerate it. Please correct me if I'm wrong, but I thought that this support will no longer be _required_ when we have a first release on the RELENG_5 (-STABLE) branch. [EMAIL PROTECTED] Cheers, -- Ruslan Ermilov Sysadmin and DBA, [EMAIL PROTECTED] Sunbay Software Ltd, [EMAIL PROTECTED] FreeBSD committer pgp0.pgp Description: PGP signature
Re: Can't build -CURRENT on 4.7
In message: [EMAIL PROTECTED] Ruslan Ermilov [EMAIL PROTECTED] writes: : On Fri, Jun 06, 2003 at 11:57:00PM -0700, David O'Brien wrote: : On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: : The compiler in 4.7 does not like this: : : -std=gnu99 : : As a result, buildworld of -CURRENT fails : rather early. : : Committers are not required to support building 5-CURRENT, post : 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the : change. However, someone will probably patch the build system to : tolerate it. : : Please correct me if I'm wrong, but I thought that this support : will no longer be _required_ when we have a first release on the : RELENG_5 (-STABLE) branch. [EMAIL PROTECTED] First off, I'd like to say that's my understanding as well. Having said that, let's not get overly anal about the rules here. There's still a great need to have current build on 4.x machines. This is a long standing range war between ruslan and david over how much compatibility should be there. I do not want to see it play out in public again, but fear that it might. Let's apply some common sense to this exceptional situation we find ourselves in and not resort to being overly pedantic about rules. There are a number of people who cannot run a 5.0-RELEASE system on their hardware because it was too instable to build a world. So requiring a trip through 5.0 is not an option. Therefore, we have to tolerate going from 4.x to at least 5.1 and maybe a little farther. We're just barely past 5.1 right now, so it is a little fast to be breaking this path. I personally don't see that the addition of -std=gnu99 is enough of a win in -current to justify its painful addition and the issue of -stable compatibility is secondary to that. It's been added 3 or 4 times now and every time the world has broken on some architecture. That alone is reason to treat the change with some skeptism as to its correctness. My main point, however, is that common sense needs to be applied to this situation, not blind adherence to inflexible rules. That's the sort of thing that gets us as a project into trouble. Warner ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
David O'Brien wrote: On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Committers are not required to support building 5-CURRENT, post 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the change. However, someone will probably patch the build system to tolerate it. Hmm.. I'll upgrade the machine to 4-STABLE and see if that addresses it. I'm also looking at at some other approaches. For example, the attached patch changes BMAKEENV to override CSTD in the early build phases. (This also required changing a couple of 'inline' to '__inline' in xlint/lint1/cgram.y.) This seems to get it through the bootstrap, at least, although I'm still running into build problems later on (but the cross-tools are built by then, so I think these may be unrelated). Tim Kientzle Index: Makefile.inc1 === RCS file: /usr/src/cvs/src/Makefile.inc1,v retrieving revision 1.363 diff -u -r1.363 Makefile.inc1 --- Makefile.inc1 31 May 2003 21:29:38 - 1.363 +++ Makefile.inc1 7 Jun 2003 04:52:43 - @@ -200,7 +204,7 @@ BMAKEENV= DESTDIR= \ INSTALL=sh ${.CURDIR}/tools/install.sh \ PATH=${BPATH}:${PATH} \ - WORLDTMP=${WORLDTMP} \ + WORLDTMP=${WORLDTMP} CSTD=c90 \ MAKEFLAGS=-m ${.CURDIR}/tools/build/mk ${.MAKEFLAGS} BMAKE= MAKEOBJDIRPREFIX=${WORLDTMP} \ ${BMAKEENV} ${MAKE} -f Makefile.inc1 \ Index: usr.bin/xlint/lint1/cgram.y === RCS file: /usr/src/cvs/src/usr.bin/xlint/lint1/cgram.y,v retrieving revision 1.7 diff -u -r1.7 cgram.y --- usr.bin/xlint/lint1/cgram.y 3 Mar 2002 15:12:19 - 1.7 +++ usr.bin/xlint/lint1/cgram.y 7 Jun 2003 06:30:12 - @@ -1642,17 +1642,17 @@ return (0); } -static inline int uq_gt(uint64_t, uint64_t); -static inline int q_gt(int64_t, int64_t); +static __inline int uq_gt(uint64_t, uint64_t); +static __inline int q_gt(int64_t, int64_t); -static inline int +static __inline int uq_gt(uint64_t a, uint64_t b) { return (a b); } -static inline int +static __inline int q_gt(int64_t a, int64_t b) { ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
On Sat, Jun 07, 2003 at 06:30:11AM -0600, M. Warner Losh wrote: In message: [EMAIL PROTECTED] Ruslan Ermilov [EMAIL PROTECTED] writes: : On Fri, Jun 06, 2003 at 11:57:00PM -0700, David O'Brien wrote: : On Fri, Jun 06, 2003 at 09:46:07PM -0700, Tim Kientzle wrote: : The compiler in 4.7 does not like this: : : -std=gnu99 : : As a result, buildworld of -CURRENT fails : rather early. : : Committers are not required to support building 5-CURRENT, post : 5.0-RELEASE on a 4.7 machine. So this is not grounds to remove the : change. However, someone will probably patch the build system to : tolerate it. : : Please correct me if I'm wrong, but I thought that this support : will no longer be _required_ when we have a first release on the : RELENG_5 (-STABLE) branch. [EMAIL PROTECTED] First off, I'd like to say that's my understanding as well. That was not my understanding at all. Having said that, let's not get overly anal about the rules here. There's still a great need to have current build on 4.x machines. This is a long standing range war between ruslan and david over how much compatibility should be there. I do not want to see it play out in public again, but fear that it might. How is this a war? My elusion to someone will probably patch the build system to tolerate it is that I expected that RU would add something to -legacy or Makefile.inc1 so that the 5-CURRENT build worked 4.0. I don't believe anyone can infer I would get in someone's way in doing this. We even have a freshly bumped __FreeBSD_version (501100) that can be used for this. I personally don't see that the addition of -std=gnu99 is enough of a win in -current to justify its painful addition and the issue of -stable compatibility is secondary to that. It's been added 3 or 4 times now and every time the world has broken on some architecture. That alone is reason to treat the change with some skeptism as to its correctness. And the TRB hasn't even responded to my or DES's emails on the topic. Note even a hi, we got your message and will mull over it. I (and another committer) had a agenda that DES's commit derailed and I'll be damned if I'm going to let it stop me given the amount of crap I've taken lately that has derailed every effort I've tried to make in FreeBSD for the past 2 months. -- -- David ([EMAIL PROTECTED]) ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
In message: [EMAIL PROTECTED] David O'Brien [EMAIL PROTECTED] writes: : First off, I'd like to say that's my understanding as well. : : That was not my understanding at all. The last time it came it, it was specifically stated that it was until the branch point. But so far the work arounds are fairly simple. I'm not going to give you more grief about it. : Having said that, let's not get overly anal about the rules here. : There's still a great need to have current build on 4.x machines. : This is a long standing range war between ruslan and david over how : much compatibility should be there. I do not want to see it play out : in public again, but fear that it might. : : How is this a war? My elusion to someone will probably patch the build : system to tolerate it is that I expected that RU would add something to : -legacy or Makefile.inc1 so that the 5-CURRENT build worked 4.0. I don't : believe anyone can infer I would get in someone's way in doing this. : We even have a freshly bumped __FreeBSD_version (501100) that can be used : for this. It isn't a war, yet. I'd like to keep it that way. There's some history here that makes me fear. However, I'll put those fears away and try to get a workaround that works. : I personally don't see that the addition of -std=gnu99 is enough of a : win in -current to justify its painful addition and the issue of : -stable compatibility is secondary to that. It's been added 3 or 4 : times now and every time the world has broken on some architecture. : That alone is reason to treat the change with some skeptism as to its : correctness. : : And the TRB hasn't even responded to my or DES's emails on the topic. : Note even a hi, we got your message and will mull over it. Noted. I'll go and re-read the trb@ request. It came in, and then there was a flurry of commits so I thought it was OBE. I'll go look into it. : I (and another committer) had a agenda that DES's commit derailed and : I'll be damned if I'm going to let it stop me given the amount of crap : I've taken lately that has derailed every effort I've tried to make in : FreeBSD for the past 2 months. I'm not trying to give you crap here. I'm trying to point out that there are additional concerns that come into play. Warner ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
On Sat, Jun 07, 2003 at 10:38:15AM -0700, Tim Kientzle wrote: Index: usr.bin/xlint/lint1/cgram.y === RCS file: /usr/src/cvs/src/usr.bin/xlint/lint1/cgram.y,v retrieving revision 1.7 diff -u -r1.7 cgram.y --- usr.bin/xlint/lint1/cgram.y 3 Mar 2002 15:12:19 - 1.7 +++ usr.bin/xlint/lint1/cgram.y 7 Jun 2003 06:30:12 - @@ -1642,17 +1642,17 @@ -static inline int uq_gt(uint64_t, uint64_t); -static inline int q_gt(int64_t, int64_t); +static __inline int uq_gt(uint64_t, uint64_t); +static __inline int q_gt(int64_t, int64_t); -static inline int +static __inline int uq_gt(uint64_t a, uint64_t b) { return (a b); } -static inline int +static __inline int Good catch! The rest of the file used __include everywhere. I've fixed the inconsistency. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
On Sat, Jun 07, 2003 at 10:38:15AM -0700, Tim Kientzle wrote: Index: Makefile.inc1 === RCS file: /usr/src/cvs/src/Makefile.inc1,v retrieving revision 1.363 diff -u -r1.363 Makefile.inc1 --- Makefile.inc1 31 May 2003 21:29:38 - 1.363 +++ Makefile.inc1 7 Jun 2003 04:52:43 - @@ -200,7 +204,7 @@ BMAKEENV=DESTDIR= \ INSTALL=sh ${.CURDIR}/tools/install.sh \ PATH=${BPATH}:${PATH} \ - WORLDTMP=${WORLDTMP} \ + WORLDTMP=${WORLDTMP} CSTD=c90 \ This won't work on non-i386, due to alloca issues. + WORLDTMP=${WORLDTMP} CSTD= \ may. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
In message: [EMAIL PROTECTED] David O'Brien [EMAIL PROTECTED] writes: : + WORLDTMP=${WORLDTMP} CSTD= \ : : may. This seems to mostly work... I'm trying to sort out a couple of other issues, but those may be related to other changes that I've made in my test tree. Warner ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
David O'Brien wrote: This won't work on non-i386, due to alloca issues. + WORLDTMP=${WORLDTMP} CSTD= \ may. Hmmm... This seems like the Right Thing in any case, since it is one less assumption you're making about the build environment. I'm still getting buildworld failures, though. Long after the bootstrap, using the new tools, I'm seeing consistent failures in libpthread: building shared library libkse.so.1 thr_libc.So: In function `sigaction': thr_libc.So(.text+0x54): multiple definition of `_sigaction' thr_sigaction.So:/usr/src/current/lib/libpthread/thread/thr_sigaction.c:43: first defined here thr_libc.So: In function `sigprocmask': thr_libc.So(.text+0x34): multiple definition of `_sigprocmask' thr_sigprocmask.So:/usr/src/current/lib/libpthread/thread/thr_sigprocmask.c:46: first defined here *** Error code 1 Stop in /usr/src/current/lib/libpthread. *** Error code 1 Tim ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Can't build -CURRENT on 4.7
The compiler in 4.7 does not like this: -std=gnu99 As a result, buildworld of -CURRENT fails rather early. Tim Kientzle ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to [EMAIL PROTECTED]