|
From: Dirk M. <dm...@gm...> - 2005-06-15 23:41:27
Attachments:
fix_regtest.diff
valgrind-glibc24.diff
|
Hi, I need the attached patches to compile valgrind 2.4.0 against gcc4 / glibc2.4. As i got burned when I tried to checkin something last time, I'm posting the patches here for review first. I'd like to apply them to valgrind-3.0 too btw. Dirk |
|
From: Julian S. <js...@ac...> - 2005-06-16 11:26:41
|
Dirk Thanks for these. I'll apply the glibc one to the 3 line. I wasn't clear what problem the fix_regtest patch solves -- can you clarify? I see it decides if gcc understands -msse/-mmmx (what about -msse2 ?) and I know that -msse2 engages gcc4's autovectoriser, but how does that help with the regtests? J On Thursday 16 June 2005 00:41, Dirk Mueller wrote: > Hi, > > I need the attached patches to compile valgrind 2.4.0 against gcc4 / > glibc2.4. > > As i got burned when I tried to checkin something last time, I'm posting > the patches here for review first. > > I'd like to apply them to valgrind-3.0 too btw. > > Dirk |
|
From: Dirk M. <dm...@gm...> - 2005-06-16 11:59:34
|
On Thursday 16 June 2005 13:26, Julian Seward wrote: > Thanks for these. I forgot the glibc-2.4.supp file btw, its still in the works. > I see it decides if gcc understands -msse/-mmmx (what about > -msse2 ?) and I know that -msse2 engages gcc4's autovectoriser, > but how does that help with the regtests? the fpu and sse regression tests contain assembler sequences which reference MMX/SSE registers. gcc 4.x does not accept usage of those registers unless explicitely told about it with -mmmx / -msse. of course we could use something like -march=i[56]86 as well, because then it knows that those registers exist on the target platform. explicitely hardcoding -march=i686 is rather bad though, if you want to compile e.g. -march=pentium-m (which is "better" than i686). Dirk |
|
From: Nicholas N. <nj...@cs...> - 2005-06-16 13:20:36
|
On Thu, 16 Jun 2005, Dirk Mueller wrote: > the fpu and sse regression tests contain assembler sequences which reference > MMX/SSE registers. gcc 4.x does not accept usage of those registers unless > explicitely told about it with -mmmx / -msse. I've been using a 4.0 pre-release (haven't bothered upgrading to 4.0.0 yet) for a while without problems. Any ideas why I haven't had a problem? N |
|
From: Julian S. <js...@ac...> - 2005-06-16 15:29:36
|
On Thursday 16 June 2005 14:20, Nicholas Nethercote wrote: > On Thu, 16 Jun 2005, Dirk Mueller wrote: > > the fpu and sse regression tests contain assembler sequences which > > reference MMX/SSE registers. gcc 4.x does not accept usage of those > > registers unless explicitely told about it with -mmmx / -msse. > > I've been using a 4.0 pre-release (haven't bothered upgrading to 4.0.0 > yet) for a while without problems. Any ideas why I haven't had a problem? Yes, me too -- I've been building the 3 line on Fedora Core 4 for a while now, which is gcc4 based, with no problems. Umm... J |
|
From: Dirk M. <dm...@gm...> - 2005-06-16 16:47:47
|
On Thursday 16 June 2005 15:20, Nicholas Nethercote wrote: > I've been using a 4.0 pre-release (haven't bothered upgrading to 4.0.0 > yet) for a while without problems. Any ideas why I haven't had a problem? I don't know. Its been in gcc 4.0 CVS branch (which has been released as 4.0.1 for example) for a while, and its also in the SUSE incarnation of gcc 4.0 CVS which I use, and so far I didn't bother to figure out. Do you really want to know? Dirk |
|
From: Julian S. <js...@ac...> - 2005-06-17 09:33:26
|
Dirk > On Thursday 16 June 2005 15:20, Nicholas Nethercote wrote: > > I've been using a 4.0 pre-release (haven't bothered upgrading to 4.0.0 > > yet) for a while without problems. Any ideas why I haven't had a > > problem? > > I don't know. Its been in gcc 4.0 CVS branch (which has been released as > 4.0.1 for example) for a while, and its also in the SUSE incarnation of gcc > 4.0 CVS which I use, and so far I didn't bother to figure out. Do you > really want to know? Sorry to be tedious, but I'd like to understand a bit more. It strikes me as strange that they would change the behaviour of the compiler in the 4.0.0 - 4.0.1 transition. (and 4.0.1 is released? gcc.gnu.org doesn't mention this). Could you send the error message(s) that you get, so I can see what it is gcc is complaining about? J |
|
From: Nicholas N. <nj...@cs...> - 2005-06-17 13:26:24
|
On Fri, 17 Jun 2005, Julian Seward wrote: > (and 4.0.1 is released? gcc.gnu.org doesn't mention this). I believe it's currently "frozen", ie. close to being released, but isn't out yet. Maybe a week or so. N |
|
From: Nicholas N. <nj...@cs...> - 2005-06-30 04:07:51
|
On Thu, 16 Jun 2005, Dirk Mueller wrote: > I need the attached patches to compile valgrind 2.4.0 against gcc4 / glibc2.4. > > As i got burned when I tried to checkin something last time, I'm posting the > patches here for review first. > > I'd like to apply them to valgrind-3.0 too btw. I've committed the entire 2nd patch, which added support for glibc-2.4 to configure.in. I also added an empty 2.4 suppression file, Dirk can you fill that in when you find time? I didn't use the first patch (except removing asm/ipc.h from scalar.c) because we didn't resolve whether the extra MMX/SSE flags were necessary. I only put them in the 3.0 repository. I don't think we'll be doing a 2.4.1 release, in which case the 2.4 repo is effectively dead and there's no point updating it. N |
|
From: Dirk M. <dm...@gm...> - 2005-07-03 02:49:44
|
On Thursday 30 June 2005 06:07, Nicholas Nethercote wrote: > I've committed the entire 2nd patch, which added support for glibc-2.4 to > configure.in. I also added an empty 2.4 suppression file, Dirk can you > fill that in when you find time? Sure. > I didn't use the first patch (except removing asm/ipc.h from scalar.c) > because we didn't resolve whether the extra MMX/SSE flags were necessary. Well, they are necessary with newer gcc's. Want me to show you the gcc changelog? or the error message? > I only put them in the 3.0 repository. I don't think we'll be doing a > 2.4.1 release, in which case the 2.4 repo is effectively dead and there's > no point updating it. I think there is a point updating it, because 2.4.0 right now doesn't even compile, so a 2.4.1 at least fixing that would be great. It has low priority though. |
|
From: Julian S. <js...@ac...> - 2005-07-03 10:43:19
|
Dirk
> > I didn't use the first patch (except removing asm/ipc.h from scalar.c)
> > because we didn't resolve whether the extra MMX/SSE flags were necessary.
>
> Well, they are necessary with newer gcc's. Want me to show you the gcc
> changelog? or the error message?
It's this I guess:
2005-05-10 Michael Matz <ma...@su...>
from trunk:
2005-04-28 James E Wilson <wi...@sp...>
* config/i386/i386.h (HI_REGISTER_NAMES): Fix typos in comment.
(ADDITIONAL_REGISTER_NAMES): Delete obsolete mmx register entries.
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/i386/i386.h.diff?r1=1.421.6.3&r2=1.421.6.4&f=h
I just find it bizarre that on a stable branch, they changed the
externally visible behaviour of the compiler. You would think they
would only do bug fixes in the 4.0.0 -> 4.0.1 transition.
But they did. So I committed your patch on the 3 line. Thanks.
J
|
|
From: Dirk M. <dm...@gm...> - 2005-07-05 20:25:17
|
On Sunday 03 July 2005 12:43, Julian Seward wrote: > I just find it bizarre that on a stable branch, they changed the > externally visible behaviour of the compiler. You would think they > would only do bug fixes in the 4.0.0 -> 4.0.1 transition. Bug fixes sometimes also involve compliance fixes ;-) Anyway, thanks for committing. the version-script-removal commit is missing on the 3.0 tree of valgrind, as well as the disable-pie-by-default change should be backported to 2.4. Can I do that? Dirk |
|
From: Nicholas N. <nj...@cs...> - 2005-07-05 20:30:20
|
On Tue, 5 Jul 2005, Dirk Mueller wrote: > Anyway, thanks for committing. the version-script-removal commit is > missing on the 3.0 tree of valgrind, as well as the > disable-pie-by-default change should be backported to 2.4. Can I do > that? Sounds fine to me. I believe Julian is planning to do a 3.0 release soon, and so it seems unlikely that a 2.4.1 release will occur, so I wouldn't bother with the backport. (Sorry I didn't point this out when you removed valgrind.vs from 2.4, I thought you did it with the 3.0 repository.) N |
|
From: Dirk M. <dm...@gm...> - 2005-07-05 21:22:49
|
On Tuesday 05 July 2005 22:30, Nicholas Nethercote wrote: > Sounds fine to me. I believe Julian is planning to do a 3.0 release soon, > and so it seems unlikely that a 2.4.1 release will occur I'd be willing to release-master a 2.4.1 if thats the only reason it won't get done.. > , so I wouldn't > bother with the backport. (Sorry I didn't point this out when you removed > valgrind.vs from 2.4, I thought you did it with the 3.0 repository.) I don't have access to 3.0 right now. May I sign up for an account? Thanks, Dirk |