* [testsuite] Provide TLS access model testcases
@ 2011-05-23 13:53 Rainer Orth
2011-05-29 16:44 ` Kaz Kojima
0 siblings, 1 reply; 6+ messages in thread
From: Rainer Orth @ 2011-05-23 13:53 UTC (permalink / raw)
To: gcc-patches; +Cc: Uros Bizjak
While working to improve Solaris/x86 TLS support, Uros provided me with
a testcase to excercise all 4 TLS access models. Since there were
different issues affecting the different models, I've split the testcase
into one per access model so one can easier see what's broken.
Tested with the appropriate runtest invocations in the whole range of
Solaris configurations, as well as Tru64 UNIX:
as, ld gas, ld gas, gld
alpha-dec-osf5.1b PASS
i386-pc-solaris2.8 PASS gd, ld FAIL PASS
i386-pc-solaris2.9 PASS gd, ld FAIL PASS
i386-pc-solaris2.10 gd, ld FAIL gd, ld FAIL PASS
i386-pc-solaris2.11 PASS PASS PASS
sparc-pc-solaris2.8 PASS PASS PASS
sparc-pc-solaris2.9 PASS PASS PASS
sparc-pc-solaris2.10 PASS PASS PASS
sparc-pc-solaris2.11 PASS PASS PASS
The Solaris 8 and 9 as, ld configurations are actually emutls, as is
Tru64 UNIX, so we won't see failures on such targets.
I do have a patch for the gd and ld failures on Solaris/x86. Those
models weren't excercised anywhere in the testsuite before. The
failures occur since the Solaris ABI requires slightly different code
sequences here, and the patch accomodates this. In the very latest
version of Solaris 11, the linker maintainers have added support for the
GNU code sequences, too, so all is fine there.
While testing the patch, the Solaris 8 and 9 gas, ld and gas, gld
configurations were UNSUPPORTED initially, since the tls_runtime test
itself failed:
ld.so.1: tls_runtime4775.exe: fatal: tls_runtime4775.exe: object requires TLS, but TLS failed to initialize
Adding the necessary TLS options fixes this and provides the results
above.
I'll install the patch in a day unless someone sees any problems.
Rainer
2010-12-30 Uros Bizjak <ubizjak@gmail.com>
Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
* gcc.dg/torture/tls/run-gd.c: New test.
* gcc.dg/torture/tls/run-ie.c: New test.
* gcc.dg/torture/tls/run-ld.c: New test.
* gcc.dg/torture/tls/run-le.c: New test.
* lib/target-supports.exp (check_effective_target_tls_runtime):
Build testcase with TLS options.
diff --git a/gcc/testsuite/gcc.dg/torture/tls/run-gd.c b/gcc/testsuite/gcc.dg/torture/tls/run-gd.c
new file mode 100644
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/tls/run-gd.c
@@ -0,0 +1,32 @@
+/* { dg-do run } */
+/* { dg-require-effective-target tls_runtime } */
+/* { dg-add-options tls } */
+
+extern void abort (void);
+
+__thread int tls_gd __attribute__((tls_model("global-dynamic"))) = 0;
+
+int get_gd (void)
+{
+ return tls_gd;
+}
+
+int *get_gdp (void)
+{
+ return &tls_gd;
+}
+
+int main (void)
+{
+ int val;
+
+ val = get_gd ();
+ if (val != 0)
+ abort ();
+
+ val = *get_gdp ();
+ if (val != 0)
+ abort ();
+
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/torture/tls/run-ie.c b/gcc/testsuite/gcc.dg/torture/tls/run-ie.c
new file mode 100644
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/tls/run-ie.c
@@ -0,0 +1,32 @@
+/* { dg-do run } */
+/* { dg-require-effective-target tls_runtime } */
+/* { dg-add-options tls } */
+
+extern void abort (void);
+
+__thread int tls_ie __attribute__((tls_model("initial-exec"))) = 4;
+
+int get_ie (void)
+{
+ return tls_ie;
+}
+
+int *get_iep (void)
+{
+ return &tls_ie;
+}
+
+int main (void)
+{
+ int val;
+
+ val = get_ie ();
+ if (val != 4)
+ abort ();
+
+ val = *get_iep ();
+ if (val != 4)
+ abort ();
+
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/torture/tls/run-ld.c b/gcc/testsuite/gcc.dg/torture/tls/run-ld.c
new file mode 100644
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/tls/run-ld.c
@@ -0,0 +1,33 @@
+/* { dg-do run } */
+/* { dg-require-effective-target tls_runtime } */
+/* { dg-add-options tls } */
+
+extern void abort (void);
+
+__thread int tls_ld __attribute__((tls_model("local-dynamic"))) = 1;
+__thread int tls_ld2 __attribute__((tls_model("local-dynamic"))) = 2;
+
+int get_ld (void)
+{
+ return tls_ld + tls_ld2;
+}
+
+int *get_ldp (void)
+{
+ return &tls_ld;
+}
+
+int main (void)
+{
+ int val;
+
+ val = get_ld ();
+ if (val != 1 + 2)
+ abort ();
+
+ val = *get_ldp ();
+ if (val != 1)
+ abort ();
+
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/torture/tls/run-le.c b/gcc/testsuite/gcc.dg/torture/tls/run-le.c
new file mode 100644
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/tls/run-le.c
@@ -0,0 +1,32 @@
+/* { dg-do run } */
+/* { dg-require-effective-target tls_runtime } */
+/* { dg-add-options tls } */
+
+extern void abort (void);
+
+__thread int tls_le __attribute__((tls_model("local-exec"))) = 3;
+
+int get_le (void)
+{
+ return tls_le;
+}
+
+int *get_lep (void)
+{
+ return &tls_le;
+}
+
+int main (void)
+{
+ int val;
+
+ val = get_le ();
+ if (val != 3)
+ abort ();
+
+ val = *get_lep ();
+ if (val != 3)
+ abort ();
+
+ return 0;
+}
--
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [testsuite] Provide TLS access model testcases
2011-05-23 13:53 [testsuite] Provide TLS access model testcases Rainer Orth
@ 2011-05-29 16:44 ` Kaz Kojima
2011-05-30 15:59 ` Rainer Orth
2011-06-07 11:54 ` Joseph S. Myers
0 siblings, 2 replies; 6+ messages in thread
From: Kaz Kojima @ 2011-05-29 16:44 UTC (permalink / raw)
To: gcc-patches; +Cc: ro, ubizjak
Hi,
Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> wrote:
> 2010-12-30 Uros Bizjak <ubizjak@gmail.com>
> Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
>
> * gcc.dg/torture/tls/run-gd.c: New test.
> * gcc.dg/torture/tls/run-ie.c: New test.
> * gcc.dg/torture/tls/run-ld.c: New test.
> * gcc.dg/torture/tls/run-le.c: New test.
> * lib/target-supports.exp (check_effective_target_tls_runtime):
> Build testcase with TLS options.
gcc.dg/torture/tls/run-gd.c fails on SH because SH assumes
-fpic for global and local dynamic tls models. Although a line
/* { dg-options "-fpic" { target sh*-*-* } } */
added to run-gd.c and run-ld.c tests looks enough for SH,
I guess that target fpic instead of target sh*-*-* is more
reasonable because GD and LD tls models are usually for shared
objects.
I see similar failures on S390 testresults and it seems that
also S390 requires -fpic for GD and LD.
Give me your thoughts?
Regards,
kaz
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [testsuite] Provide TLS access model testcases
2011-05-29 16:44 ` Kaz Kojima
@ 2011-05-30 15:59 ` Rainer Orth
2011-05-31 7:53 ` Kaz Kojima
2011-06-07 11:54 ` Joseph S. Myers
1 sibling, 1 reply; 6+ messages in thread
From: Rainer Orth @ 2011-05-30 15:59 UTC (permalink / raw)
To: Kaz Kojima; +Cc: gcc-patches, ubizjak
Kaz Kojima <kkojima@rr.iij4u.or.jp> writes:
> gcc.dg/torture/tls/run-gd.c fails on SH because SH assumes
> -fpic for global and local dynamic tls models. Although a line
>
> /* { dg-options "-fpic" { target sh*-*-* } } */
>
> added to run-gd.c and run-ld.c tests looks enough for SH,
> I guess that target fpic instead of target sh*-*-* is more
> reasonable because GD and LD tls models are usually for shared
> objects.
I don't think so: on all my targets (Solaris/SPARC and /x86, IRIX and
Tru64 UNIX), the test works without -fpic, so this seems to be quite
target specific.
> I see similar failures on S390 testresults and it seems that
> also S390 requires -fpic for GD and LD.
> Give me your thoughts?
Given that this seems to be target-specific, unless we can identify a
common feature of these targets, I'd rather see us xfail or skip the
tests on a per-target basis.
In addition, I've got a patch in the works that cycles the
gcc.dg/torture/tls tests through -fpic/-fPIC and -fpie/-fPIE if
available. This way, the testcases will be exercised on sh*-*-*, too.
Rainer
--
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [testsuite] Provide TLS access model testcases
2011-05-30 15:59 ` Rainer Orth
@ 2011-05-31 7:53 ` Kaz Kojima
0 siblings, 0 replies; 6+ messages in thread
From: Kaz Kojima @ 2011-05-31 7:53 UTC (permalink / raw)
To: ro; +Cc: gcc-patches, ubizjak
Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> wrote:
> Given that this seems to be target-specific, unless we can identify a
> common feature of these targets, I'd rather see us xfail or skip the
> tests on a per-target basis.
>
> In addition, I've got a patch in the works that cycles the
> gcc.dg/torture/tls tests through -fpic/-fPIC and -fpie/-fPIE if
> available. This way, the testcases will be exercised on sh*-*-*, too.
Sounds good. I'll add a line like
/* { dg-skip-if "" { sh*-*-* } { "*" } { "-fpic" "-fPIC" "-fpie" "-fPIE" } } */
to run-[gl]d.c when your patch is installed.
Thank you for your suggestion.
Regards,
kaz
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [testsuite] Provide TLS access model testcases
2011-05-29 16:44 ` Kaz Kojima
2011-05-30 15:59 ` Rainer Orth
@ 2011-06-07 11:54 ` Joseph S. Myers
2011-06-07 13:35 ` Kaz Kojima
1 sibling, 1 reply; 6+ messages in thread
From: Joseph S. Myers @ 2011-06-07 11:54 UTC (permalink / raw)
To: Kaz Kojima; +Cc: gcc-patches, ro, ubizjak
On Sun, 29 May 2011, Kaz Kojima wrote:
> Hi,
>
> Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE> wrote:
> > 2010-12-30 Uros Bizjak <ubizjak@gmail.com>
> > Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
> >
> > * gcc.dg/torture/tls/run-gd.c: New test.
> > * gcc.dg/torture/tls/run-ie.c: New test.
> > * gcc.dg/torture/tls/run-ld.c: New test.
> > * gcc.dg/torture/tls/run-le.c: New test.
> > * lib/target-supports.exp (check_effective_target_tls_runtime):
> > Build testcase with TLS options.
>
> gcc.dg/torture/tls/run-gd.c fails on SH because SH assumes
> -fpic for global and local dynamic tls models. Although a line
That sounds like an SH target bug to me.
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [testsuite] Provide TLS access model testcases
2011-06-07 11:54 ` Joseph S. Myers
@ 2011-06-07 13:35 ` Kaz Kojima
0 siblings, 0 replies; 6+ messages in thread
From: Kaz Kojima @ 2011-06-07 13:35 UTC (permalink / raw)
To: joseph; +Cc: gcc-patches, ro, ubizjak
"Joseph S. Myers" <joseph@codesourcery.com> wrote:
>> gcc.dg/torture/tls/run-gd.c fails on SH because SH assumes
>> -fpic for global and local dynamic tls models. Although a line
>
> That sounds like an SH target bug to me.
Ok, I'd like to fix the SH backend so that those tests pass
without -fpic/-fpie.
Regards,
kaz
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2011-06-07 13:35 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-05-23 13:53 [testsuite] Provide TLS access model testcases Rainer Orth
2011-05-29 16:44 ` Kaz Kojima
2011-05-30 15:59 ` Rainer Orth
2011-05-31 7:53 ` Kaz Kojima
2011-06-07 11:54 ` Joseph S. Myers
2011-06-07 13:35 ` Kaz Kojima
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).