https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101081
Bug ID: 101081
Summary: analyzer testsuite failures seen with new glibc due to
malloc attribute
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: analyzer
Assignee: dmalcolm at gcc dot gnu.org
Reporter: dmalcolm at gcc dot gnu.org
Target Milestone: ---
Fedora's build of GCC showed some new analyzer failures:
+FAIL: gcc.dg/analyzer/analyzer-verbosity-2a.c (test for excess errors)
+FAIL: gcc.dg/analyzer/analyzer-verbosity-3a.c (test for excess errors)
+FAIL: gcc.dg/analyzer/edges-1.c (test for excess errors)
+FAIL: gcc.dg/analyzer/file-1.c (test for excess errors)
+FAIL: gcc.dg/analyzer/file-2.c (test for excess errors)
+FAIL: gcc.dg/analyzer/file-paths-1.c (test for excess errors)
+FAIL: gcc.dg/analyzer/file-pr58237.c (test for excess errors)
+FAIL: gcc.dg/analyzer/pr99716-1.c (test for excess errors)
which appeared between:
gcc/11.1.1/3.fc35:
glibc-2.33.9000-2.fc35.x86_64
gcc/11.1.1/4.fc35:
glibc-2.33.9000-13.fc35.x86_64
These testcases all use <stdio.h>
Almost all of these seem to be due to new -Wanalyzer-malloc-leak warnings,
where the testcases are expecting to emit -Wanalyzer-file-leak (from
sm-file.cc), but are now *also* emitting -Wanalyzer-malloc-leak.
Presumably the new glibc gained __attribute__((malloc)) within stdio.h.
As well as these, pr99716-1.c adds these excess errors:
/builddir/build/BUILD/gcc-11.1.1-20210531/gcc/testsuite/gcc.dg/analyzer/pr99716-1.c:11:5:
warning: use of possibly-NULL 'fp' where non-null expected [CWE-690]
[-Wanalyzer-possible-null-argument]
/builddir/build/BUILD/gcc-11.1.1-20210531/gcc/testsuite/gcc.dg/analyzer/pr99716-1.c:23:5:
warning: use of possibly-NULL 'fp' where non-null expected [CWE-690]
[-Wanalyzer-possible-null-argument]
/builddir/build/BUILD/gcc-11.1.1-20210531/gcc/testsuite/gcc.dg/analyzer/pr99716-1.c:36:5:
warning: use of possibly-NULL '*fpp' where non-null expected [CWE-690]
[-Wanalyzer-possible-null-argument]
which look like genuine warnings: passing unchecked ptrs to fprintf.
Looks like:
(a) these testcases should avoid <stdio.h> and instead have their own copy of
the decls, to avoid <stdio.h> from changing under us
(b) the malloc attribute gave us some genuine warnings, so it's valuable; the
analyzer should do something to support this, but it's probably overkill to
emit *both* -Wanalyzer-malloc-leak and -Wanalyzer-file-leak.
