fix lock and tls macros

2007-11-04 Thread Bruno Haible
It's necessary to protect the statements to which the lock and tls macros expand against use in an 'if' branch. test-tls.c was not compiled the way it was intented on mingw because an 'else' keyword was connected to the wrong 'if' statement. This should fix it. 2007-11-04 Bruno Haible <[EMAIL P

Re: new module 'nocrash'

2007-11-04 Thread Bruno Haible
Benoit SIGOURE wrote: > I've found this: > http://www.nirsoft.net/vc/ > prevent_application_crash_exception_handling.html > but haven't had time to test it so far. An alternative is to use SetUnhandledExceptionFilter; see libsigsegv-2.4/src/handler-win32.c for how to do it. Bruno

make use of 'nocrash' in printf tests

2007-11-04 Thread Bruno Haible
This makes use of the 'nocrash' module to avoid a dialog box from the gl_PRINTF_ENOMEM configure check. Tested on MacOS X 10.3.9: After executing the command $ defaults write com.apple.CrashReporter DialogType crashreport I get crash reports. (The values can be different for other OS versions: t

Re: new module 'nocrash'

2007-11-04 Thread Benoit SIGOURE
On Nov 4, 2007, at 7:50 PM, Bruno Haible wrote: Hi, Some autoconf tests require running a program and seeing whether it crashes or not. Unfortunately, one some platforms, this causes a dialog box to appear. If a configure script has 10 test programs of this kind, the installer can mutate

new module 'nocrash'

2007-11-04 Thread Bruno Haible
Hi, Some autoconf tests require running a program and seeing whether it crashes or not. Unfortunately, one some platforms, this causes a dialog box to appear. If a configure script has 10 test programs of this kind, the installer can mutate into a click-o-maniac; in other words, it's annoying. On

YAPPP: yet another *printf portability problem

2007-11-04 Thread Bruno Haible
The test cases for large precisions, meant to uncover out-of-memory handling problems, also uncovered another problem: - On mingw, precisions larger than 512 are capped, i.e. treated like if they were 512. So, you ask for 2000 digits, you get 512 digits. - On BeOS, precisions larger than 10

vasnprintf fix

2007-11-04 Thread Bruno Haible
The floating-point output code crashed due to an abort() for values with large exponents (> 1e34 for 'double'). This fixes it. 2007-11-04 Bruno Haible <[EMAIL PROTECTED]> * lib/vasnprintf.c (scale10_round_decimal_decoded): Fix shift loop. --- lib/vasnprintf.c.orig 2007-11-04 14:5

Re: AC_LIB_PROG_LD

2007-11-04 Thread Sylvain Beucler
On Sun, Nov 04, 2007 at 10:33:32AM +0100, Bruno Haible wrote: > Sylvain Beucler wrote: > > relocatable.m4 is using the AC_LIB_PROG_LD definition. > > AC_REQUIRE([AC_LIB_PROG_LD])dnl we use $LD > > > > However this macro is defined in gettext > > (autoconf-lib-link/m4/lib-ld.m4), so i

Re: AC_LIB_PROG_LD

2007-11-04 Thread Bruno Haible
Sylvain Beucler wrote: > relocatable.m4 is using the AC_LIB_PROG_LD definition. > AC_REQUIRE([AC_LIB_PROG_LD])dnl we use $LD > > However this macro is defined in gettext > (autoconf-lib-link/m4/lib-ld.m4), so it's not available in a minimal > autoconf + automake install. Fixed like