On Tue, May 22, 2012 at 4:57 PM, Roland Mainz <[email protected]> wrote:
> On Mon, May 21, 2012 at 3:20 PM, Glenn Fowler <[email protected]> wrote:
>>
>> On Sun, 20 May 2012 21:24:15 +0200 Irek Szczesniak wrote:
>>> On Sun, May 20, 2012 at 9:11 PM, Irek Szczesniak <[email protected]>
>>> wrote:
>>> > On Fri, May 18, 2012 at 9:35 PM, Glenn Fowler <[email protected]>
>>> > wrote:
>>> >> the AT&T Software Technology ast beta 2012-05-18 source release
>>> > On Suse I can't get a 32bit build (/bin/ksh ./bin/package make
>>> > PACKAGE_OPTIONS=map-libc CCFLAGS='-DSHOPT_CMDLIB_BLTIN=0 -DSHOPT_SYSRC
>>> > -D_map_libc=1 -g -ggdb -fstrict-aliasing -Wstrict-aliasing' CC='gcc
>>> > -std=gnu99 -m32 -fPIC' HOSTTYPE=linux.i386 SHELL=/bin/ksh) pass *any*
>>> > of the tests. I get a huge row of memory corruption failures:
>>> This is odd: The failure occurs only for ast-open builds. If I build
>>> ast-ksh it does not occur.
>>
>> one difference is .a's vs. shared libs
>> the ast-ksh build does not have nmake so it does not build shared libs
>> but since for most systems ksh links with .a ast libraries its probably
>> a build-time rather than run-time link issue
>> I'll try to reproduce today
>
> Uhm... this _may_ actually be one of my old nemesises (sorry for
> forgetting this one) when building ast-open. AFAIK the problem is that
> something in the build system goes "mad" and randomly enables
> SHOPT_*-flags (AFAIK in this case this may be |SHOPT_SPAWN|).
> Somewhere on my list of forgotten things (that's why I was praying for
> a bugzilla to document such bugs and work on them later) is an item
> which says "... test mam*-tools and nmake against valgrind and
> Rational Purify... " ... I bet there is a simple variable without a
> defined value floating around which causes this.
Mhhh... I have to defer the testing. A quick attempt (e.g. $ CC='gcc
-std=gnu99 -g' CCFLAGS='-g' valgrind --trace-children=yes
--track-origins=yes --read-var-info=yes
"--log-file=$PWD/valgrindlogs/%p.log" ~/bin/ksh ./bin/package make
2>&1 | tee buildlog.log #) failed... it seems the combination
"valgrind vs. /bin/package make" doesn't work well because suddenly
all hell breaks loose (and runs around):
-- snip --
+ /bin/cp dbm.req
/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/arch/linux.i386-64/lib/lib/dbm
/bin/cp: cannot create regular file
`/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/arch/linux.i386-64/lib/lib/dbm':
No such file or directory
mamake [cmd/INIT]: *** exit code 1 making
/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/arch/linux.i386-64/lib/lib/dbm
mamake: *** exit code 1 making cmd/INIT
+
/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/arch/linux.i386-64/bin/mamake
-C lib/libast -k install
probing C language processor
/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/arch/linux.i386-64/bin/cc
for mam information
/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/arch/linux.i386-64/bin/cc:
not a C compiler: failed to compile ``int x;''
+ iffe -v -X ast -X std -c 'cc ${mam_cc_DLL} -D_BLD_ast -g ' run
/home/test001/work/ast_ksh_20120518/build_normal_32bit_mamfilefix_valgrind/src/lib/libast/features/standards
iffe: cc ${mam_cc_DLL} -D_BLD_ast -g : not a C compiler: should
compile 'int i = 1;'
mamake [lib/libast]: *** exit code 1 making FEATURE/standards
-- snip --
I'll look at this again after the next beta is out...
----
Bye,
Roland
--
__ . . __
(o.\ \/ /.o) [email protected]
\__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer
/O /==\ O\ TEL +49 641 3992797
(;O/ \/ \O;)
_______________________________________________
ast-developers mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-developers
