[Bug other/51450] New: configure's test for -fno-rtti -fno-exceptions broken

[Bug other/51450] New: configure's test for -fno-rtti -fno-exceptions broken

[christophe.lyon at st dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51450

             Bug #: 51450
           Summary: configure's test for -fno-rtti -fno-exceptions broken
    Classification: Unclassified
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: christophe.lyon@st.com


I have noticed that configure decides that my build GCC does not support
-fno-rtti -fno-exceptions when building a new GCC.

I traced this down to a problem in libtool.m4 where _LT_COMPILER_NO_RTTI uses
_LT_COMPILER_OPTION to check whether these 2 options are supported.

The latter compiles & links $lt_simple_compile_test_code which contains:
lt_simple_compile_test_code="int some_variable = 0;"

The link phase fails because there is no main(), and then configure decides
-fno-rtti -fno-exceptions are not supported.

I am not sure about how to fix this:
- change _LT_COMPILER_OPTION to use $lt_simple_link_test_code instead?
- use a modified _LT_LINKER_OPTION (given that the options tests belong to
CFLAGS rather than LDFLAGS)?
- other?

Maybe this bug should be reported elsewhere?

[Bug bootstrap/51450] configure's test for -fno-rtti -fno-exceptions broken

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51450

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2012-01-28
     Ever Confirmed|0                           |1

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-01-28 19:41:20 UTC ---
The configure which is the subversion on the trunk has the correct int main(){}
test in it.

Are you regenerating configure at all?

[Bug bootstrap/51450] configure's test for -fno-rtti -fno-exceptions broken

[christophe.lyon at st dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51450

--- Comment #2 from christophe.lyon at st dot com 2012-01-30 09:23:27 UTC ---
I noticed this about an official release; I am not sure which one it was when I
entered this report, but it is at least true for GCC-4.6.2.

I have just checked
http://gcc.gnu.org/viewcvs/trunk/gcc/configure?revision=183406&view=markup
which seems to be the current trunk version of GCC's configure and can still
see:
lt_simple_compile_test_code="int some_variable = 0;"
at line 14904

So, when this is used when testing for -fno-rtti -fno-exceptions, it makes the
test fail.

I am not regenerating configure.

[Bug bootstrap/51450] configure's test for -fno-rtti -fno-exceptions broken

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51450

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-01-30 20:18:31 UTC ---
Looks like there are two different places in configure which test -fno-rtti and
-fno-exceptions:
One around line 28658 (which does not use lt_simple_compile_test_code
And one around line 14958 (which does use it).

I was looking at the first one rather than the second one.

[Bug bootstrap/51450] configure's test for -fno-rtti -fno-exceptions broken

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51450

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW

[Bug bootstrap/51450] configure's test for -fno-rtti -fno-exceptions broken

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51450

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
configure:18495: checking if gcc supports -fno-rtti -fno-exceptions
configure:18513: gcc -c -g   -fno-rtti -fno-exceptions conftest.c >&5
cc1: warning: command-line option '-fno-rtti' is valid for C++/D/ObjC++ but not
for C
configure:18517: $? = 0
configure:18530: result: no


It should have used g++ here ...

Note the previous one works though:
configure:7500: checking whether gcc supports -fno-rtti
configure:7517: gcc -c -fno-rtti  conftest.c >&5
cc1: warning: command-line option '-fno-rtti' is valid for C++/D/ObjC++ but not
for C
configure:7517: $? = 0
configure:7526: result: yes




https://lists.gnu.org/archive/html/libtool/2003-08/msg00003.html

Looks like the issue upstream in libtool now the question comes is this fixed
upstream or not.

Note the test here is not an issue for GCC's builds in general due to the extra
testing for -fno-rtti -fno-exceptions that gcc does but still an issue for the
libtool that is included with gcc.