[Bug regex/28170] New: attribute access (write_only, 4, 3) on regexec triggers a warning

["msebor at gmail dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=28170

            Bug ID: 28170
           Summary: attribute access (write_only, 4, 3) on regexec
                    triggers a warning
           Product: glibc
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: regex
          Assignee: unassigned at sourceware dot org
          Reporter: msebor at gmail dot com
                CC: drepper.fsp at gmail dot com
  Target Milestone: ---

Commit 26492c0a14966c32c43cd6ca1d0dca5e62c6cfef
(https://sourceware.org/pipermail/glibc-cvs/2021q2/073049.html) adds attribute
access to a number of APIs, including the regexec function like so:

@@ -648,10 +655,12 @@ extern int regcomp (regex_t *_Restrict_ __preg,
 extern int regexec (const regex_t *_Restrict_ __preg,
                    const char *_Restrict_ __String, size_t __nmatch,
                    regmatch_t __pmatch[_Restrict_arr_],
-                   int __eflags);
+                   int __eflags)
+    __attr_access ((__write_only__, 4, 3));

The attribute was added here based on the POSIX description of the function
that describes the pmatch array referenced by it as write-only (at least
nothing I can find in POSIX says the function writes to it).

With a recent GCC 12 enhancement to detect read accesses to write-only objects
the presence of the attribute triggers the following warning:

regexec.c: In function ‘__regexec’:
regexec.c:204:13: warning: ‘*pmatch.rm_so’ is used uninitialized
[-Wuninitialized]
regexec.c:192:101: note: accessing argument 4 of a function declared with
attribute ‘access (write_only, 4, 3)’

The warning indicates that the attribute may not, in fact, appropriate for the
function when since it writes to the referenced object.  The write takes place
when the REG_STARTEND bit (a Glibc extension) is set in flags.

-- 
You are receiving this mail because:
You are on the CC list for the bug.