Hi, 
 
I tried this with gtk from 1-2 and it fails. 
 
I tried building this with pkg-config and gtk+-2.0 and I get the following compiler error: 
 
/src2/ooo11rc/crashrep/source/unx/mai n.cxx  
/src2/ooo11rc/crashrep/source/unx/main.cxx:65:22: _version.h: No such file or  
directory /src2/ooo11rc/crashrep/source/unx/main.cxx:97: warning: ISO C++  
forbids declaration of `g_b DebugMode' with no type  
/src2/ooo11rc/crashrep/source/unx/main.cxx: In function `bool  
write_report(const _STL::hash _map<_STL::string, _STL::string,  
_STL::hash<_STL::string>, _STL::equal_to<_STL::string>, _S  
TL::allocator<_STL::pair<const _STL::string, _STL::string> > >&)':  
/src2/ooo11rc/crashrep/source/unx/main.cxx:337: `_INPATH' undeclared (first  
use this functi                                                                 
on)  
 
So either way, this will not build. 
 
Why isn't this using vcl? 
 
Kevin 

Hi, 
 
Three other things: 
 
1. pkg-config gives --only-mod-libs: unknown option 
 
 
2. I changed crashrep/source/unx/main.cxx  
 
#include <_version.h> to instead be 
#include <vcl/_version.h>  
 
since there is no _version.h in the include path 
 
 
3. The link command ends up with -W,  
with nothing after the comma which messes up the linker causing the final link to fail 
 
This happens because ... 
 
.IF "$(ENABLE_STATIC_GTK)" == "FALSE" 
GTKLINKFLAGS= 
.ELSE 
GTKLINKFLAGS=-Bstatic 
.ENDIF 
 
GTKLINKFLAGS is set to nothing but 
 
APP1STDLIBS=-Wl,$(GTKLINKFLAGS) ... 
 
So this has to be hacked away too. 
 
With all of these changes I can get it to build but ... 
 

Hi, 
 
Some other things that should probably to be done for crash reporter: 
 
1. fix declarations to be C style and not C++ to allow it to compile with -ansi 
(from Pavel) 
 
--- signal.c.~1.13.~    2003-06-18 23:50:15.000000000 +0200 
+++ signal.c    2003-06-21 16:28:15.000000000 +0200 
@@ -327,12 +327,14 @@ 
                                void *stackframes[1024]; 
                                int  iFrame; 
                                int  nFrames = backtrace( stackframes, 
sizeof(stackframes)/sizeof(stackframes[0])); 
+                               FILE *xmlout; 
+                               FILE *stackout; 
                         
                                pXMLTempName = tmpnam( szXMLTempNameBuffer ); 
                                pStackTempName = tmpnam( szStackTempNameBuffer ); 
  
-                               FILE *xmlout = fopen( pXMLTempName, "w" ); 
-                               FILE *stackout = fopen( pStackTempName, "w" ); 
+                               xmlout = fopen( pXMLTempName, "w" ); 
+                               stackout = fopen( pStackTempName, "w" ); 
  
                                fprintf( xmlout, "<errormail:Stack type=\"%s\">\n", STACKTYPE 
); 
  
@@ -447,11 +449,10 @@ 
 static void PrintStack( int sig ) 
 { 
        void *buffer[MAX_FRAME_COUNT]; 
+       int size = backtrace( buffer, sizeof(buffer) / sizeof(buffer[0]) ); 
  
        fprintf( stderr, "\n\nFatal exception: Signal %d\n", sig ); 
  
-       int size = backtrace( buffer, sizeof(buffer) / sizeof(buffer[0]) ); 
- 
        if ( size > 0 ) 
        { 
                fputs( "Stack:\n", stderr ); 
 
 
2. handle case of 0 return from dladdr so that null stirng is not sent to strchr 
 
>       dladdr( stackframes[iFrame], &dl_info); 
> 
>       fprintf(stderr,"reached point 2\n");fflush(stderr); 
> 
>      dli_fname = strrchr(  dl_info.dli_fname, '/' ); 
 
> But there is no guarantee that dlinfo.dli_fname is not null  
> which in turn can generate a segfault in strchr  
> so need to check return vale of dladdr. 
 
3. set sigaction so that signal handler can block other signals on entry 
    to prevent recursion 
 
4. change jre to no_jre in SOLARLIB to prevent libjpeg.so from JDK 1.3.1 builds 
    from being linked into crash_report 
 
>SOLARLIB!:=$(SOLARLIB:s/jre/no_jre/) 
 
 5. define a BUILD_SPECIAL style macro so that SERVER= is filled in our 
bootstraprc 
 
  #ifdef OSL_PRODUCT 
>    #ifdef BUILD_SPECIAL 
>      Value  = "report.services.openoffice.org"; 
>    #else 
>      Value  = ""; 
>    #endif 
> 
 

Hi, 
 
Adding myself to CC on this 
 
Kevin 
 

signal.c is already fixed.

Hi, 
 
Did you add the fix for testing return value from dladdr in signal.c or is that still 
needed? 
 
Kevin 
 

../../unxlngi4.pro/obj/main.o(.text+0x178): In function
`write_stack(long)':
: the use of `tmpnam' is dangerous, better use `mkstemp'


This must be removed or else it will conflict with packaging rules and
cuase problems.

@Kevin: no, only C++ vs. C changes.

Created attachment 7057 [details]
short patch to replace tmpnam with mkstemp and to handle null dladdr fields

Created attachment 7058 [details]
changes to crash_rep makefiles and main.cxx

Created attachment 7064 [details]
a better signal_fix patch that changes fewer lines

Hi, 
 
I did not like my first signal_fix.txt patch since it did not proerly handle 
the ifdef case when no backtrace was desired. 
 
I also changed to many lines when a simpler version was possible. 
 
So please use new_signal_fix.txt as a patch file and not signal_fix.txt  
if you want things to build with mkstemp and handle the dladdr issue. 
 
Hope this helps, 
 
Kevin 
 

Hi,

--- source/unx/main.cxx.~1.6.~	2003-06-20 13:08:59.000000000 -0400
+++ source/unx/main.cxx	2003-06-20 13:08:59.000000000 -0400
@@ -62,7 +62,7 @@
 #include <interface.hxx>
 #include <cstdio>
 #include <sys/utsname.h>
-#include <_version.h>
+#include <vcl/_version.h>

This didn't work for me.
Because vcl was not built before crashrep in my environment,
kernel-2.2.18, glibc-2.1.3, gcc-3.3, XFree86-4.1.0, jdk-1.4.1
 

Hi, 
 
As a workaround, simply add vcl as a dependency on the top line of 
crashrep/prj/build.lst 
 
cr      crashrep        :       sal sysui gtk tools NULL 
 
In fact since we don't have a gtk module we should be replacing gtk with vcl  
 
This should allow the build to continue until a better fix is found. 
 
I certainly hope we are not going to add the mishmash that is gtk+2 to our build 
requirements or our source base. 
 
Kevin 
 

I'll take over this one.

Target set.

Fixed non failsafe dladdr usage in 

sal/osl/unx/signal.c branch cws_srx645_ooo11rc rev 1.13.4.2

Use mkstemp instead of tmpnam in
sal/osl/unx/signal.c branch cws_srx645_ooo11rc rev 1.13.4.4

Include <vcl/_version.h> to get _INPATH define and other in
crashrep/source/unx/main.cxx branch cws_srx645_ooo11rc rev 1.6.6.2

Fixed linker command line if static gtk is NOT used in
crashrep/source/unx/makefile.mk branch cws_srx645_ooo11rc rev 1.2.6.1

Didn't used the suggested fixed because they introduced some new errors.

@Pavel: Please review and check if it works.

@Kevin: Don't change the build lst in module crashrep as this will
break SO build dependencies. Just ignore the warning of a missing gtk
module until we have a bette solution.

Hi, 
 
Reopening this. 
 
Fixes are missing in crashrep/source/unx/makefile.mk 
 
--only-mod-libs does NOT exist in current pkg-config  
 
The following line needs to be added to both crashrep/source/unx/makefile.mk and 
crashrep/source/unx/res/makefile.mk 
 
SOLARLIB!:=$(SOLARLIB:s/jre/no_jre/)  
 
Otherwise during link SOLARLIB will find libjpeg.so from JDK 1.3.1 which has other 
pieces linked into it causing problems. 
 
Also, none of these changes have been committed to OOo 1.1RC yet. 
 
Kevin 
 

Done in branch cws_srx645_ooo11rc. Will automaticly be commited to OOo
1.1RC by merging the branch by tomorrow.

Hi, 
 
Looks great! 
 
Thanks! 
 
Kevin 
   

Hi, 
 
cvs update shows all fixes in place so  verified. 
 
Kevin 
 

We've still got the --only-mod-libs problem for
ENABLE_STATIC_GTK=TRUE.  Without this fixed I can't test linking
statically to GTK.

crashrep/source/unx$ ENABLE_STATIC_GTK=TRUE dmake
------------------------------
Making: ../../unxlngi4.pro/bin/crash_report.bin
unx
--only-mod-libs: unknown option
dmake:  Error code 1, while making
'../../unxlngi4.pro/bin/crash_report.bin'
---* TG_SLO.MK *---

JA->HRO: Martin mentioned that the remaining problem of Chris might be
a duplicate of an issue which you're working on ...? Reassigning to you

Please refrain from building crashrep with static linked gtk libraries
and forget 'configure --use-static-gtk'.
Without the pkg-config option '--only-mod-libs' some common libraries
like dl etc. will be linked statically too. That would require
additional changes to pkg-config. I was not aware that the pkg-config
here differs from that available on a scratch OOo build.

It's a better option to ship the needed GTK libraries (LGPL) with OOo
and always link dynamically.
For OOo 1.1 RC all builds for Linux x86 and Solaris will be done by
SUN so we can provide a correct static linked version for download.

I've submitted an issue 16528 to do this for OOo 1.1.
Sorry for the inconvenience.

Hi,  
  
> It's a better option to ship the needed GTK libraries (LGPL) with OOo  
> and always link dynamically.  
> For OOo 1.1 RC all builds for Linux x86 and Solaris will be done by  
> SUN so we can provide a correct static linked version for download.  
>   
  
This is a joke right?  You really want to add the gtk-+2.0 libs to the binary tarball  
people have to download (and all of the supporting libs too?).    
  
That seems completely crazy to add such a build requirement for anything like a  
crash reporter.  
  
Please simply rewrite crashreporter to use a X11, Xt and Xaw and drop all this idea  
of including gtk+2.0 libs in the installer.  
  
PPC Linux simply will not be shipping any gtk-2.0 libs and therefore probably not  
crashreporter in the future either then.  
  
I will instead ifdef the code in sal/osl/unx/signal.c to rename the created stack trace  
tmp files to include something more amenable to users with full date and ask the user  
to forward those files with any crashing bug report.  
  
Kevin  
 
  
  

closed.