Public note of failure (apparently since test case was added): With LAST_UPDATED "Sat Nov 6 19:25:03 GMT 2004", gcc/testsuite/gcc.c-torture/compile/20031023-1.c: In function 'bar': gcc/testsuite/gcc.c-torture/compile/20031023-1.c:38: internal compiler error: in change_address_1, at emit-rtl.c:1821
Hmm, I cannot reproduce this on a cross from powerpc-darwin (but it might be a local patch which is causing it to pass which I doubt it).
Fresh checkout on different system on i686-pc-linux-gnu (FC2).
... LAST_UPDATED "Mon Nov 8 00:53:12 UTC 2004", that is.
A bug in the epilogue code: using "int" instead of HOST_WIDE_INT. This is a regression from 3.2, where the code got a clear error (the test-case wasn't present there, though): /gcc/testsuite/gcc.c-torture/compile/20031023-1.c: In function `bar': /gcc/testsuite/gcc.c-torture/compile/20031023-1.c:38: stack frame too big A fix seems trivial.
This isn't a 4.0 regression, it's a 3.2 regression. If you change it again, please add a comment.
if you mean it is a regression from 3.2, then to mark it as such, not the other way around which is what you marked it as.
In reply to Comment #6, I did in comment #4. I also changed the PR title, because [3.2 regression] meant to me a regression from 3.2. I see <URL:http://gcc.gnu.org/bugs/management.html> says otherwise. Oh well. Though as I'm not going to change 3.3 or 3.4 (too odd case, but do feel free after a build and check), I changed back the title to say 4.0.
...oh right: "needs fixing" isn't the same as "going to be fixed". Changed back to 3.3/3.4/4.0. Gah.
Subject: Bug 18321 CVSROOT: /cvs/gcc Module name: gcc Changes by: hp@gcc.gnu.org 2004-12-28 23:19:13 Modified files: gcc : ChangeLog gcc/config/mmix: mmix.c Log message: PR target/18321 * config/mmix/mmix.c (mmix_expand_epilogue): Change type of variable offset to HOST_WIDE_INT. Remove obsolete comment. Fix spacing. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.6973&r2=2.6974 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/mmix/mmix.c.diff?cvsroot=gcc&r1=1.72&r2=1.73
<URL:http://gcc.gnu.org/ml/gcc-patches/2004-12/msg02020.html>
Hans, this was also a regression in 3.3 and 3.4, the patch should be applied there, or you should clarify that you are not going to fix the branches.
In response to comment #11: You should clarify what clarification you expect besides what I say in comment #7 and the target milestone setting. Perhaps you missed it, because it was said in passing? In <URL:http://gcc.gnu.org/bugs/management.html> i see nothing further on this. Please add whatever clarification there too. Again: I'm not going to apply the patch myself on those branches, but I preapprove backporting it there, assuming the required before/after test is done, of course.
Hans: our policy prescribes that a regression in Bugzilla be kept open until it is fixed on all the active release branches, *or* the maintainter of the code affected by the bug explicitally decides that this bug will not be fixed in the branches. This said, I had missed your note in comment #7. Since you are the maintainer of mmix, it is up to you to decide if you want to keep this bug in the branches or fix it there as well. If you prefer to not fix it, then it's fine.
Subject: Re: [3.3/3.4/4.0 regression] mmix-knuth-mmixware testsuite failure: gcc.c-torture/compile/20031023-4.c -O0, -O1 "giovannibajo at libero dot it" <gcc-bugzilla@gcc.gnu.org> writes: | Hans: our policy prescribes that a regression in Bugzilla be kept open until it | is fixed on all the active release branches, *or* the maintainter of the code | affected by the bug explicitally decides that this bug will not be fixed in the | branches. | | This said, I had missed your note in comment #7. Since you are the maintainer | of mmix, it is up to you to decide if you want to keep this bug in the branches | or fix it there as well. If you prefer to not fix it, then it's fine. If HP believs that he has no interest in having the bug fixed on the branches for mmix-knuth-xx users, let's close the PR. -- Gaby
> If HP believs that he has no interest in having the bug fixed on the > branches for mmix-knuth-xx users, let's close the PR. I have interest, but only a little, not even as much as doing the minimal work myself. The PR is already closed. Nothing here to see, move on.