This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: simple question about testresults list
- To: Vincent Diepeveen <diep at xs4all dot nl>
- Subject: Re: simple question about testresults list
- From: DJ Delorie <dj at redhat dot com>
- Date: 10 Feb 2001 12:12:47 -0500
- Cc: gcc at gcc dot gnu dot org
- References: <3.0.32.20010210060234.00b549b0@pop.xs4all.nl>
Vincent Diepeveen <diep@xs4all.nl> writes:
> Hello, how to read the below testresults in the testresults list, as
> i don't understand the whole line at all, are there .c files that by
> accident carries a date less as a month ago?
In the gcc sources, there is a directory gcc/testsuite. In there are
many small source files that are used to verify that the compiler is
working correctly. In the examples below, for example, one of the
files is gcc/testsuite/gcc.c-torture/compile/20010113-1.c
The files are named according to the date they're added to the
testsuite, so 20010115-2.c would be January 15th, 2001, and the second
one added that day.
Recent cases are more likely to fail than older ones because there has
been less time for someone to come up with a fix for them.
> Also nearly all optimization flags it fails upon this .c file:
> 20010113-1.c
That means that the problem isn't in the optimizer, then. Problems
that involve the optimizer usually fail only on some of the
combinations. Problems outside the optimizer usually fail everything.
> How bad is it to fail like that on this file, is the .c file simply
> wrong?
It depends on the file, what it's testing, and how often you use code
like that in your own constructs. Some of the tests are pretty basic,
others are very subtle.
It's rare that a test case is wrong, but not impossible.