https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107750
--- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot
Uni-Bielefeld.DE> ---
> --- Comment #2 from David Malcolm <dmalcolm at gcc dot gnu.org> ---
> Sorry about the breakage; thanks for filing this, and for the patch.
>
> Looking at:
> FAIL: gcc.dg/analyzer/fd-access-mode-target-headers.c (test for warnings,
> line 53)
> it looks like the analyzer was able to figure out the values of O_RDONLY and
> O_WRONLY in your headers, but not of O_ACCMODE. I suspect it might be having
> issues with SOCK_STREAM and SOCK_DGRAM also.
>
> How are these values defined in your headers? The analyzer has gained some
> ability to try to e.g. scrape macro values from the C frontend, but only some
> cases are supported.
<sys/fcntl.h>:
#define O_ACCMODE (O_SEARCH | O_EXEC | 0x3)
<sys/socket.h>:
#if !defined(_XPG4_2) || defined(__EXTENSIONS__)
#define SOCK_STREAM NC_TPI_COTS /* stream socket */
#define SOCK_DGRAM NC_TPI_CLTS /* datagram socket */
#define SOCK_RAW NC_TPI_RAW /* raw-protocol interface */
#else
#define SOCK_STREAM 2 /* stream socket */
#define SOCK_DGRAM 1 /* datagram socket */
#define SOCK_RAW 4 /* raw-protocol interface */
#endif /* !defined(_XPG4_2) || defined(__EXTENSIONS__) */
NC_TPI_COTS is also 2, same for the other two.
> There's a way of debugging this using -fdump-analyzer, but I'm looking at
> adding a way to make such issues show up more clearly in the DejaGnu logs,
> which hopefully should make them easier to track down.
That would certainly be nice.
Thanks.
Rainer
