This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/68273] [5/6 Regression] Wrong code on mips/mipsel due to (invalid?) peeking at alignments in function_arg.
- From: "law at redhat dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Fri, 05 Feb 2016 17:45:33 +0000
- Subject: [Bug target/68273] [5/6 Regression] Wrong code on mips/mipsel due to (invalid?) peeking at alignments in function_arg.
- Auto-submitted: auto-generated
- References: <bug-68273-4 at http dot gcc dot gnu dot org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68273
--- Comment #19 from Jeffrey A. Law <law at redhat dot com> ---
Created attachment 37599
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=37599&action=edit
Testcase for the testsuite
My best attempt at pulling together the right dg-markers for this test.
mips.exp does some "interesting" stuff. I think I managed to narrow it down to
the case we want to test and reasonable strings to search for in the RTL
expansion dump.
Essentially it verifies that we've got a single assignment to $6 (for the
3-argument call to sprintf) and two assignments to $5 (one for the sprintf call
one for the enter call).
When we generate the wrong code we have two assignments to $6 and one to $5.
Use as you see fit.