https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120591
Bug ID: 120591
Summary: sarif tests depend on (length of) directory name
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: testsuite
Assignee: unassigned at gcc dot gnu.org
Reporter: gjl at gcc dot gnu.org
Target Milestone: ---
When doing regression tests, I found that a sarif test depends on the directory
name of builddir, which leads to extra extra noise in the testsuite results:
# Comparing directories
## Dir1=/home/gjl/gnu/build/gcc-15-avr/gcc/testsuite/: 1 sum files
## Dir2=/home/gjl/gnu/build/gcc-15-x/gcc/testsuite/: 1 sum files
# Comparing 1 common sum files
## /bin/sh /home/gjl/gnu/source/gcc-master/contrib/compare_tests
/tmp/gxx-sum1.1055378 /tmp/gxx-sum2.1055378
New tests that FAIL (1 tests):
atmega128-sim: gcc: gcc.dg/sarif-output/bad-pragma.c
bad-pragma.py::test_messages_have_embedded_urls - assert "unknown
opti...s.html)' ..
Old tests that failed, that have disappeared (1 tests): (Eeek!)
atmega128-sim: gcc: gcc.dg/sarif-output/bad-pragma.c
bad-pragma.py::test_messages_have_embedded_urls - assert "unknown
opti...s.html)' ki..
## Differences found
# 1 differences in 1 common sum files found
Notice the trailing ".." vs. "ki.." in the output, which makes
contrib/compare_tests to report changes in the testsuite results when there are
acually none. The respective part of the diff of gcc.sum looks like this:
$ diff ~/gnu/build/gcc-15-*/gcc/testsuite/gcc/gcc.sum
11c11
< Running
/home/gjl/gnu/source/gcc-15/gcc/testsuite/gcc.dg/sarif-output/sarif-output.exp
...
---
> Running
> /home/gjl/gnu/source/gcc-15-x/gcc/testsuite/gcc.dg/sarif-output/sarif-output.exp
> ...
15c15
< FAIL: gcc.dg/sarif-output/bad-pragma.c
bad-pragma.py::test_messages_have_embedded_urls - assert "unknown
opti...s.html)' ki..
---
> FAIL: gcc.dg/sarif-output/bad-pragma.c
> bad-pragma.py::test_messages_have_embedded_urls - assert "unknown
> opti...s.html)' ..
A diagnostic should not depend on some directory names like builddir.
