Here is propoused patch which checks if demangle.h is present.
Darrell, I have demangle.h on my system, so it's up to you to test it.
Actually anybody can test --- just temporarily rename or move /usr/include/demangle.h. :-)
Darrell
2013/8/16 Darrell Anderson darrella@hushmail.com
Here is propoused patch which checks if demangle.h is present.
Darrell, I have demangle.h on my system, so it's up to you to test it.
Actually anybody can test --- just temporarily rename or move /usr/include/demangle.h. :-)
The previous patch makes turning on the debug quite tricky, so better to remove the ifdef's with NDEBUG in backtrace_symbols.c.
I've tested the build both with and without demangle.h and seems everything is ok. so retest
Dne pá 16. srpna 2013 Fat-Zer napsal(a):
2013/8/16 Darrell Anderson darrella@hushmail.com
Here is propoused patch which checks if demangle.h is present.
Darrell, I have demangle.h on my system, so it's up to you to test it.
Actually anybody can test --- just temporarily rename or move /usr/include/demangle.h. :-)
The previous patch makes turning on the debug quite tricky, so better to remove the ifdef's with NDEBUG in backtrace_symbols.c.
I've tested the build both with and without demangle.h and seems everything is ok. so retest
It was necessary to add a condition for linking with BFD. Updated patch attached. Please test it.
Slavek --
2013/8/16 Slávek Banko slavek.banko@axis.cz
Dne pá 16. srpna 2013 Fat-Zer napsal(a):
2013/8/16 Darrell Anderson darrella@hushmail.com
Here is propoused patch which checks if demangle.h is present.
Darrell, I have demangle.h on my system, so it's up to you to test it.
Actually anybody can test --- just temporarily rename or move /usr/include/demangle.h. :-)
The previous patch makes turning on the debug quite tricky, so better to remove the ifdef's with NDEBUG in backtrace_symbols.c.
I've tested the build both with and without demangle.h and seems everything is ok. so retest
It was necessary to add a condition for linking with BFD. Updated patch attached. Please test it.
Slavek
Sorry, I'm a little confused, so, slackware doesn't have libbfq at all or just demangle.h is missing? If the matter is about linking than IMHO the solution with conditional user-defined option is better.
Anyway in a day or two I'll came with a normal patch to move all that to kdebug.cpp and provide WITH_LIBBFD for optional linkage... + some additional backtrace features.
On Friday 16 of August 2013 20:15:31 Fat-Zer wrote:
2013/8/16 Slávek Banko slavek.banko@axis.cz
Dne pá 16. srpna 2013 Fat-Zer napsal(a):
2013/8/16 Darrell Anderson darrella@hushmail.com
Here is propoused patch which checks if demangle.h is present.
Darrell, I have demangle.h on my system, so it's up to you to test it.
Actually anybody can test --- just temporarily rename or move /usr/include/demangle.h. :-)
The previous patch makes turning on the debug quite tricky, so better to remove the ifdef's with NDEBUG in backtrace_symbols.c.
I've tested the build both with and without demangle.h and seems everything is ok. so retest
It was necessary to add a condition for linking with BFD. Updated patch attached. Please test it.
Slavek
Sorry, I'm a little confused, so, slackware doesn't have libbfq at all or just demangle.h is missing? If the matter is about linking than IMHO the solution with conditional user-defined option is better.
Anyway in a day or two I'll came with a normal patch to move all that to kdebug.cpp and provide WITH_LIBBFD for optional linkage... + some additional backtrace features.
When there is no demangle.h => not used full backtrace => is unnecessary linking libbfd. On Debian / Ubuntu is demangle.h along with libbfd.so contained in the package binutils-dev. If binutils-dev is not installed, demangle.h is not present and also not be linking to libbfd. Therefore, this condition is necessary.
Add WITH_LIBBFD choice is in my opinion nonsense. In this case, it makes more sense to me to add choice WITH_TDEIOSLAVE_BACKTRACE or generally WITH_BACKTRACE_HANDLER.
Therefore I propose to use either:
1) automatic enabling backtrace handler by HAVE_BACKTRACE and HAVE_DEMANGLE_H => patch from http://trinity-devel.pearsoncomputing.net/?0::11361
2) add choise WITH_TDEIOSLAVE_BACKTRACE (or WITH_BACKTRACE_HANDLER) to enable backtrace handler => patch from http://trinity-devel.pearsoncomputing.net/?0::11354
Because, unlike WITH_KDESKTOP_LOCK_BACKTRACE, in this case it was a temporary solution for debugging, I personally prefer the first option.
What is your opinion?
Slavek
On Saturday 17 of August 2013 01:31:47 Slávek Banko wrote:
On Friday 16 of August 2013 20:15:31 Fat-Zer wrote:
2013/8/16 Slávek Banko slavek.banko@axis.cz
Dne pá 16. srpna 2013 Fat-Zer napsal(a):
2013/8/16 Darrell Anderson darrella@hushmail.com
> Here is propoused patch which checks if demangle.h is present.
Darrell, I have demangle.h on my system, so it's up to you to test it.
Actually anybody can test --- just temporarily rename or move /usr/include/demangle.h. :-)
The previous patch makes turning on the debug quite tricky, so better to remove the ifdef's with NDEBUG in backtrace_symbols.c.
I've tested the build both with and without demangle.h and seems everything is ok. so retest
It was necessary to add a condition for linking with BFD. Updated patch attached. Please test it.
Slavek
Sorry, I'm a little confused, so, slackware doesn't have libbfq at all or just demangle.h is missing? If the matter is about linking than IMHO the solution with conditional user-defined option is better.
Anyway in a day or two I'll came with a normal patch to move all that to kdebug.cpp and provide WITH_LIBBFD for optional linkage... + some additional backtrace features.
When there is no demangle.h => not used full backtrace => is unnecessary linking libbfd. On Debian / Ubuntu is demangle.h along with libbfd.so contained in the package binutils-dev. If binutils-dev is not installed, demangle.h is not present and also not be linking to libbfd. Therefore, this condition is necessary.
Add WITH_LIBBFD choice is in my opinion nonsense. In this case, it makes more sense to me to add choice WITH_TDEIOSLAVE_BACKTRACE or generally WITH_BACKTRACE_HANDLER.
Therefore I propose to use either:
- automatic enabling backtrace handler by HAVE_BACKTRACE and
HAVE_DEMANGLE_H => patch from http://trinity-devel.pearsoncomputing.net/?0::11361
- add choise WITH_TDEIOSLAVE_BACKTRACE (or WITH_BACKTRACE_HANDLER) to
enable backtrace handler => patch from http://trinity-devel.pearsoncomputing.net/?0::11354
Because, unlike WITH_KDESKTOP_LOCK_BACKTRACE, in this case it was a temporary solution for debugging, I personally prefer the first option.
What is your opinion?
Slavek
I decided as follows: Now I push a patch from option 1, to resolve acute FTBFS. The advantage is that there is no need for additional build option that should be changed prospectively. And integration with the kdebug, which is a good idea, it can be done later.
Slavek
2013/8/17 Slávek Banko slavek.banko@axis.cz
I decided as follows: Now I push a patch from option 1, to resolve acute FTBFS. The advantage is that there is no need for additional build option that should be changed prospectively. And integration with the kdebug, which is a good idea, it can be done later.
Slavek
yes, sure, go ahead... I'll provide my work as I finish.
FIO: There is an another FTBFS with recent binutils e.g. 2.23.1. See upstream bug 14243: http://sourceware.org/bugzilla/show_bug.cgi?id=14243
2013/8/17 Fat-Zer fatzer2@gmail.com
2013/8/17 Slávek Banko slavek.banko@axis.cz
I decided as follows: Now I push a patch from option 1, to resolve acute FTBFS. The advantage is that there is no need for additional build option that should be changed prospectively. And integration with the kdebug, which is a good idea, it can be done later.
Slavek
yes, sure, go ahead... I'll provide my work as I finish.
FIO: There is an another FTBFS with recent binutils e.g. 2.23.1. See upstream bug 14243: http://sourceware.org/bugzilla/show_bug.cgi?id=14243
*I meant FYI, sorry.
-
Darrell Anderson -
Fat-Zer -
Slávek Banko