[Bug target/37624] New: [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[Bug target/37624] New: [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[hp at gcc dot gnu dot org]

With revision 140565 this test passed.
>From revision 140575 and on, this test has failed as follows:

Running
/tmp/hpautotest-gcc1/gcc/libstdc++-v3/testsuite/libstdc++-dg/conformance.exp
...
FAIL: 22_locale/num_get/get/char/10.cc execution test

With the message in the logfile being:

assertion "err == ios_base::failbit" failed: file
"/tmp/hpautotest-gcc1/gcc/libstdc++-v3/testsuite/22_locale/num_get/get/char/10.cc",
line 63, function: void test01()^M
program stopped with signal 6.^M

Oddly enough, I don't see this failure for other targets at gcc-testresults@ so
perhaps it's a target bug.  Of course, it could still be a non-target bug like
an uninitialized variable.  I'm labelling it as a target bug pending further
investigation.

Anyway, author of patches in suspect revision range CC:ed.


-- 
           Summary: [4.4 Regression]: 22_locale/num_get/get/char/10.cc
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hp at gcc dot gnu dot org
  GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: cris-axis-elf


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug target/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #1 from paolo dot carlini at oracle dot com  2008-09-23 16:12 -------
I'm certainly willing to help but please provide more info: e.g., what is the
value of err? Which is the path followed by cris-axis-elf in
__convert_to_v(const char* __s, long double& __v, ...)?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug target/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #2 from paolo dot carlini at oracle dot com  2008-09-23 16:24 -------
Ok, I think I know what's going wrong: please confirm that your target is using
sscanf.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug target/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #3 from paolo dot carlini at oracle dot com  2008-09-23 16:29 -------
Assuming the weak check on sscanf return value is the issue, please test the
attached patch.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug target/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #4 from paolo dot carlini at oracle dot com  2008-09-23 16:30 -------
Created an attachment (id=16395)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=16395&action=view)
Draft


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug target/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #5 from paolo dot carlini at oracle dot com  2008-09-23 16:36 -------
Never mind, I know what's going wrong but the patch is not ok.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug target/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #6 from paolo dot carlini at oracle dot com  2008-09-23 16:45 -------
Unfortunately, I think we have to "xfail" the test for targets not having
available strtold and falling back to sscanf for long double parsing. Note that
DR 23 mandates, for a reason, a behaviour equivalent to strtold, *not* sscanf,
because the latter is weaker as regards error checking, essentially. On many
targets, "6e " is successfully parsed by sscanf as 6, whereas with strtold we
can tell that not all the input is consumed and issue an error from the C++
library.


-- 

paolo dot carlini at oracle dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |paolo dot carlini at oracle
                   |dot org                     |dot com
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2008-09-23 16:45:43
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug libstdc++/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo at gcc dot gnu dot org]

------- Comment #7 from paolo at gcc dot gnu dot org  2008-09-23 17:13 -------
Subject: Bug 37624

Author: paolo
Date: Tue Sep 23 17:12:26 2008
New Revision: 140602

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=140602
Log:
2008-09-23  Paolo Carlini  <paolo.carlini@oracle.com>

        PR libstdc++/37624
        * testsuite/22_locale/num_get/get/char/10.cc: Skip long double case
        when strtold is not available.
        * testsuite/22_locale/num_get/get/wchar_t/10.cc: Likewise.

Modified:
    trunk/libstdc++-v3/ChangeLog
    trunk/libstdc++-v3/testsuite/22_locale/num_get/get/char/10.cc
    trunk/libstdc++-v3/testsuite/22_locale/num_get/get/wchar_t/10.cc


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug libstdc++/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[paolo dot carlini at oracle dot com]

------- Comment #8 from paolo dot carlini at oracle dot com  2008-09-23 17:15 -------
.


-- 

paolo dot carlini at oracle dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624

[Bug libstdc++/37624] [4.4 Regression]: 22_locale/num_get/get/char/10.cc

[hp at gcc dot gnu dot org]

------- Comment #9 from hp at gcc dot gnu dot org  2008-09-23 17:20 -------
(In reply to comment #2)
> Ok, I think I know what's going wrong: please confirm that your target is using
> sscanf.

You're too fast for me! :)
Being a newlib cross target, it doesn't use strtold; newlib has no generic
strtold. See the newlib stuff in the else-clause of the "if $GLIBCXX_IS_NATIVE;
then" in configure.ac.

(There's a strtod in newlib, and long double equals double for this and many
targets.  Not an excuse for not having a C99 standard-named function though.)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37624