Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Stephan == Stephan Bergmann sberg...@redhat.com writes: Stephan Great, thanks. (So your idea would be to, in a second step, teach Stephan ABRT to issue additional gdb commands besides backtrace in case the Stephan trace contains __gnu_cxx::__verbose_terminate_handler, right?) If that GCC bug is fixed, then the existing bt full would capture the type information. The info would appear in a mangled form, but it is trivial to run that through c++filt. Like in my example: t = 0x601060 _ZTIi@@CXXABI_1.3 - $ c++filt _ZTIi typeinfo for int Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Tom I think it may be fixable in GCC. I filed a GCC bug: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57487 You can CC yourself on it if you want to see what happens. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Stephan == Stephan Bergmann sberg...@redhat.com writes: Stephan It would be cool if there were a way to see that message in the gdb Stephan backtrace. Like __gnu_cxx::__verbose_terminate_handler assembling the Stephan message and then calling a not-optimized-away helper function with the Stephan message as argument, which in turn calls fputs and abort (though I Stephan notice that __verbose_terminate_handler currently assembles messages Stephan through multiple calls to fputs, which saves it from malloc hassles). I think it may be fixable in GCC. Right now the issue is that 't' is optimized away; but in a local build I can see: (gdb) info local terminating = true t = 0x601060 _ZTIi@@CXXABI_1.3 ... which, while not completely readable, is at least transformable with c++filt or maint demangle. I pinged a local GCC hacker, I'll let you know if he thinks this is fixable. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: claims of python unit test un-debugability considered somewhat exaggerated
David == David Ostrovsky d.ostrov...@gmx.de writes: David The only question is now, how to skip the manual sourcing of David libpython.py? If i understand it right, it happens automagically on David Fedora? Yes, Fedora installs /usr/lib/debug/usr/lib64/libpython2.7.so.1.0.debug-gdb.py (e.g.) as a hook file for gdb to load. And, it sets up gdb's trusted auto-load path to allow this. See: http://sourceware.org/gdb/current/onlinedocs/gdb/Python-Auto_002dloading.html#Python-Auto_002dloading Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: claims of python unit test un-debugability considered somewhat exaggerated
Michael == Michael Stahl mst...@redhat.com writes: CCing David Malcolm. Michael (gdb) py-bt I just wanted to mention here that Phil Muldoon is working on a frame filter patch series that is basically pretty printing for stack frames. With this in place you'll no longer need a special command -- with appropriate support from Python, you'll be able to see interpreted frames interleaved with the low-level C frames. This series is nearing final review now and should appear in gdb 7.7, later this year. I think we're planning to rewrite the existing py-bt code into frame filter form ourselves. Michael unfortunately one thing that i haven't found how to do is setting a Michael breakpoint in the python code from gdb... the best i've found is a lame Michael workaround of inserting a statement into the python code to send Michael yourself a signal, which is, well, a lame workaround: os.kill(os.getpid(), signal.SIGUSR1) Michael i wonder if it would be possible to easily add something to gdb to make Michael the python debugging a bit more seamless? Eventually we'd like to have more full support for mixed interpreter and C debugging. However this is a ways off. Meanwhile, you can do it, if you know some details about the interpreter implementation. It is just a question of setting a breakpoint at the right spot in the C code, with the right condition so that it only fires for your particular bit of code. Maybe David can say how. Perhaps the function__entry SDT probe point is apropos. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Tom I think it isn't possible in general. When an exception is thrown, I Tom think all that can really be determined is the next catch point. Michael This would be perfect. Explicit rethrows are relatively rare in Michael our code, however it is completely normal to have code that we Michael simply don't understand and/or follow - whereby we do: Unfortunately, it is even worse than I made it out to be. The next catch point also includes the spots where unwinding pauses to invoke destructors. So, at the lowest level you can't easily associate a throw with a catch that you would see in the source. You can only see to spot that calls the next destructor. If you're interested you can see this in action by breaking at the libgcc:unwind probe point (or equivalently _Unwind_DebugHook but then you have to do argument decoding by hand), then disassem $_probe_arg1. That will show you the assembly for where you're about to unwind to. (catch catch and catch throw hook into somewhat higher-level exception functions in libstdc++, which is why they don't stop at destructors; but this code doesn't know about the details and just defers to the lower-level unwinder.) I thought a bit about whether we could fix the lower levels to expose the information we'd like, but I couldn't think of a good way. Tom catch catch [REGEXP] Tom catch throw [REGEXP] Tom catch rethrow [REGEXP] Michael This is really nice; the ability to hide many of the two Michael dozen+ expected exceptions in some way be really Michael useful. Unfortunately these are often of quite generic types Michael :-) Yeah, that makes it harder. If they carry any identifying markers, you can use two of the features in tandem to filter. For example: catch throw TheExceptionType if $_exception.field == 23 This would filter by type and then do some additional checking of the identity of the exception. If they don't carry identifying markers -- if it were me, I guess I would add something to make debugging simpler. Michael Tor suggested on IRC some way of ignoring specific exception throwing Michael sites which tend to creep into the code over time and need tackling. Michael Being able to say: catch throw ignore - which would ignore the last Michael thrown exception site would be really lovely ;-) This sort of thing is reasonably easy to do in Python. Fedora has had a $_caller_is function in its gdb for a while (not sure why we haven't upstreamed this yet, probably just nobody got around to it yet). This function was actually my motivation for getting into gdb hacking and working on Python in gdb -- I wanted to be able to filter breakpoints according to caller, without a lot of hassle. I've appended it for convenience. I'll file a bug to remind us to upstream this. You would use it like: catch throw if !$_caller_is('some_thrower_to_ignore') You could wrap this up in a Python command to add !$_caller_is... to existing breakpoint conditions, or a LibreOffice-specific catch throw-like command which pre-ignores uninteresting call sites, or etc. Michael Is any of that useful ? - really looking forward to getting a Michael new libstdc++ etc. with your fixes in a few months :-) An optimist, I see :) Actually, I assume all the patches will go in reasonably soon, but there's a pretty long lag until the new stuff is deployed all over, unless you're willing to build your gcc. Please feel free to CC me on gdb questions and suggestions. I'm very interested in your feedback. Tom # Caller-is functions. # Copyright (C) 2008 Free Software Foundation, Inc. # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 3 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program. If not, see http://www.gnu.org/licenses/. import gdb import re class CallerIs (gdb.Function): Return True if the calling function's name is equal to a string. This function takes one or two arguments. The first argument is the name of a function; if the calling function's name is equal to this argument, this function returns True. The optional second argument tells this function how many stack frames to traverse to find the calling function. The default is 1. def __init__ (self): super (CallerIs, self).__init__ (caller_is) def invoke (self, name, nframes = 1): frame = gdb.selected_frame () while nframes 0: frame = frame.older () nframes = nframes - 1 return frame.name () == name.string () class CallerMatches (gdb.Function): Return True if
Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Tom == Tom Tromey tro...@redhat.com writes: Tom The various low-level exception-related functions, like __cxa_throw, Tom treat the exception object as a void *. However, the value of this Tom seems to change depending on the throw point. It's clear that this Tom can't always be the argument to throw, due to scalar and object throws. Tom So I wonder what exactly it refers to. I'll have to dig a bit deeper to Tom see how all this code really works. [...] Tom It seems like it would be nice if gdb exposed some kind of convenience Tom variable so that catch catch and catch throw could be conditional on Tom the thrown object without needing the libstdc++ debuginfo. [...] Tom This may require some libstdc++ change, perhaps a probe point. I did some more digging here and wrote a few patches. When throwing an exception, the compiler arranges to allocate an internal exception object with enough extra space for the exception passed to throw. Then it copy-constructs from the thrown object into this space and it records the object's type_info into the internal exception object. I added some SDT probes to libstdc++ to expose this information more nicely (there's really no good way to do it in all cases right now, even with debuginfo installed, as a couple of the probes are mid-function). So now I can: (gdb) catch throw Catchpoint 1 (throw) (gdb) r Starting program: /home/tromey/Space/SecondArcher/build/gdb/testsuite/gdb.cp/exception [... loads of gunk ...] Catchpoint 1 (exception thrown), __cxxabiv1::__cxa_throw (obj=0x601090, tinfo=0x600e60 typeinfo for int@@CXXABI_1.3, dest=0x0) at ../../../../gcc/libstdc++-v3/libsupc++/eh_throw.cc:63 63PROBE2 (throw, obj, tinfo); (gdb) p $_exception $1 = 13 Note that the exception variable automatically has the right type: (gdb) up #1 0x00400873 in foo (i=20) at ../../../archer/gdb/testsuite/gdb.cp/exception.cc:28 28 throw (int) 13; So, cool. The bad news is, since this requires a libstdc++ patch, even once I get everything tidied up and approved and committed, it is going to be a while before you can use it, unless you're willing to build your own gcc and gdb. Tom If we had the convenience variable mentioned above, and if LibreOffice Tom has a relatively simple exception identity measure (e.g., if you only Tom throw pointers, you can just compare them with ==), then it could Tom perhaps be done by: break at the losing catch, make a conditional catch Tom throw, then re-run. This does turn out to be a tricky bit. gdb generally mimics the source language, so things like: cond 5 $_exception == 23 ... will fail if some exception thrown is not actually comparable to 23. I'm investigating some options here. Maybe a more-magical === operator, or maybe a Python convenience function like: cond 5 $_dwim_equals ($_exception, 23) I also implemented a way to filter exception catches by name: catch catch [REGEXP] catch throw [REGEXP] catch rethrow [REGEXP] That will help the above problem a bit, you can do: catch catch int if $_exception == 23 I think all this should help with the problems that started this thread. Insight, advice, ideas -- send them my way. thanks, Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Michael == Michael Meeks michael.me...@suse.com writes: Michael The basic debugging experience in these an exception broke Michael something flows is that we get an exception thrown that Michael ultimately ends up in a pathalogical situation - an abort, or Michael some similar horrible badness. At that point the most Michael interesting thing is not the catcher - which usually ends up Michael being utterly random - but the last guy that threw the Michael exception. So then as Lubos says comes the knotty job of trying Michael to put a breakpoint on the -one- exception that ends up being Michael caught where we are now [ and that of course requires Michael re-running, and inevitably we throw dozens of exceptions in the Michael normal case ]. Thanks. This kind of discussion is very helpful to me. This problem is a bit tricky. The various low-level exception-related functions, like __cxa_throw, treat the exception object as a void *. However, the value of this seems to change depending on the throw point. It's clear that this can't always be the argument to throw, due to scalar and object throws. So I wonder what exactly it refers to. I'll have to dig a bit deeper to see how all this code really works. Anyway, this makes tracking backward from std::terminate to the original throw point more difficult. It helps a bit to install the libstdc++ debuginfo. Then at least you can dig into some of the details. However, due to optimization, even with the improvements in newer version of gcc, this turns out to be less than perfect. I tried this out to see what it was like. It is kind of awful! At least, I had to dig around through several frames of libstdc++ to find the object that lead to std::terminate being called: terminate called after throwing an instance of 'char const*' Program received signal SIGABRT, Aborted. 0x003be3036285 in __GI_raise (sig=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:64 64return INLINE_SYSCALL (tgkill, 3, pid, selftid, sig); (gdb) up #1 0x003be3037b9b in __GI_abort () at abort.c:91 91raise (SIGABRT); (gdb) up #2 0x003be80bbc5d in __gnu_cxx::__verbose_terminate_handler () at ../../../../libstdc++-v3/libsupc++/vterminate.cc:95 95 abort(); (gdb) up #3 0x003be80b9e16 in __cxxabiv1::__terminate (handler=optimized out) at ../../../../libstdc++-v3/libsupc++/eh_terminate.cc:40 40handler (); (gdb) up #4 0x003be80b9e43 in std::terminate () at ../../../../libstdc++-v3/libsupc++/eh_terminate.cc:50 50__terminate (__terminate_handler); (gdb) up #5 0x003be80b9f3e in __cxxabiv1::__cxa_throw (obj=0x601090, tinfo=optimized out, dest=optimized out) at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:83 83std::terminate (); At this point I can do: (gdb) catch throw Catchpoint 1 (throw) (gdb) cond 1 obj == 0x601090 (gdb) r The program being debugged has been started already. Start it from the beginning? (y or n) y [Inferior 25299 exited] Starting program: /tmp/r warning: failed to reevaluate condition for breakpoint 1: No symbol obj in current context. warning: failed to reevaluate condition for breakpoint 1: No symbol obj in current context. warning: failed to reevaluate condition for breakpoint 1: No symbol obj in current context. Catchpoint 1 (exception thrown), __cxxabiv1::__cxa_throw (obj=0x601090, tinfo= 0x600a60, dest=0) at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:70 70header-exc.unexpectedHandler = __unexpected_handler; Ignore the warnings; I'm not sure what they are about, but I will file a bug. ... but all this still fails if you insert a manual re-throw like throw x; into the call chain. At that point it gets really messy :( Michael Really nice ! though of course - having a full stack trace Michael would make that very substantially more useful. This is reasonably easy to implement. It may be expensive. I've appended a version that does this... well, it lists file name and line number for all the frames. If you want to get a really full stack trace, capturing the arguments and locals, then you would have to do more work. Michael Even better than this would (perhaps) be a break inside thrower that Michael is caught here type breakpoint - that we could invoke to land us in Michael whatever code is going to throw as it does that [ and before it started Michael all the magic cleanup / unwinding work ]. That is - assuming that it's Michael possible for the code to know (at that point) where it will ultimately Michael end up (? ;-) I think it isn't possible in general. When an exception is thrown, I think all that can really be determined is the next catch point. What this means is that if you have a series of throws and re-throws, winding up at some catch, then the best you could do is stop at the re-throw that leads to that catch. Does that make sense? Like: void doit() { throw hi; } void dd2() { try {
Re: Gdb support for exceptions (Re: using backtrace() in exception throwing?)
Lubos == Lubos Lunak l.lu...@suse.cz writes: Tom Is there something we could do to improve it? Lubos I don't know how much control gdb over exception handling has, Lubos so I don't know :). :-) FWIW we have the same problem in reverse: the gdb group at Red Hat is, among other things, tasked with improving the C++ debugging experience. However, most of us don't actually debug C++ programs on a regular basis. We do know some issues, via bugzilla and other discussions, but I feel sure we are also missing things. Lubos What I was refering to was the problem that if a catch block Lubos catches an exception, it's often difficult to find out where it Lubos actually came from. Using 'catch catch' doesn't show where it Lubos originated (unless I missed a non-obvious way). And if the Lubos exception propagated out of complex nesting of function calls, Lubos then 'catch throw' may trigger a number of times for exceptions Lubos that will be handled elsewhere. Solving this in general looks tricky to me. I am not really sure how to do it, but I will think about it some more. The basic issue is that if 'catch throw' triggers multiple times for the same exception, then it seems that there must be code that catches the exception and then throws it again: try { } catch (blah) { throw blah; } As opposed to a true re-throw: try { } catch (blah) { throw; } AFAIK re-throws are currently not caught, see http://sourceware.org/bugzilla/show_bug.cgi?id=12824 I'm not sure whether it is possible to easily detect whether throw x is throwing some object which has already been thrown. Hopefully I'm misunderstanding the problem :) Meanwhile, I did whip up a quick-and-dirty Python-based approach. It adds a new track-throws command. This command installs a breakpoint that records the point of the most recent throw. Then you can examine the result with info last-throw. Here it is in action: (gdb) source track_throw.py (gdb) track-throws Breakpoint 1 at 0x400910 (gdb) catch catch Catchpoint 2 (catch) (gdb) run [...] Catchpoint 2 (exception caught), __cxxabiv1::__cxa_begin_catch ( exc_obj_in=0x602070) at ../../../../libstdc++-v3/libsupc++/eh_catch.cc:41 41 { (gdb) info last-throw Last exception thrown at file ../../../archer/gdb/testsuite/gdb.cp/nextoverthrow.cc, line 36 track-throws makes the breakpoint it installs user-visible so you can disable the feature simply by deleting the breakpoint. I'm curious to know if this is useful to you. Tom import gdb last_sal = None throw_bp = None class ThrowTracker(gdb.Breakpoint): def __init__(self): gdb.Breakpoint.__init__(self, '__cxa_throw') def stop(self): global last_sal frame = gdb.newest_frame().older() last_sal = frame.find_sal() return False class TrackThrows(gdb.Command): def __init__(self): gdb.Command.__init__(self, 'track-throws', gdb.COMMAND_BREAKPOINTS) def invoke(self, arg, from_tty): global throw_bp if throw_bp is None or not throw_bp.is_valid(): # Still no good way to create a pending breakpoint from # Python. save = gdb.parameter('breakpoint pending') gdb.execute('set breakpoint pending on', to_string = True) throw_bp = ThrowTracker() if save is None: arg = 'auto' elif save: arg = 'on' else: arg = 'off' gdb.execute('set breakpoint pending %s' % arg, to_string = True) class InfoThrow(gdb.Command): def __init__(self): gdb.Command.__init__(self, 'info last-throw', gdb.COMMAND_BREAKPOINTS) def invoke(self, arg, from_tty): global last_sal if last_sal is not None: filename = last_sal.symtab.filename line = last_sal.line print Last exception thrown at file %s, line %d % (filename, line) else: print No previous exception seen TrackThrows() InfoThrow() ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: using backtrace() in exception throwing?
Lubos == Lubos Lunak l.lu...@suse.cz writes: Lubos This could be very useful ('catch throw' is so cumbersome in Lubos gdb), Is there something we could do to improve it? Lubos but I think the first thing to check is if it would Lubos actually work. Last time I checked, recent features like Lubos -fvisibility make backtrace() very often generate call traces Lubos that are next to useless. There are a few other approaches that are possible on Linux. There is one that is part of GCC, libbacktrace. It is its own library, but part of the GCC tree; so you'd have to extract it. Jan is also writing one based on elfutils. It is in elfutils git but not merged to master yet. Also it can be done using the existing unwinder data, which already exists for exception handling. This is the approach libgcj took. Specifically, it used this data to construct the list of PC values and stash this in the exception; this was then transformed when actually printing an exception. The code in libgcj is a bit hairier than you might need, since it had to handle interpreter and libffi frames specially. You could probably adapt it. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: using backtrace() in exception throwing?
Noel == Noel Grandin noelgran...@gmail.com writes: Noel Would be nice if gdb could actually do catch throw XXXException like Noel the documentation says - perhaps the documentation was modified Noel prematurely? Yeah, I think it was. The docs have been changed to reflect reality. Actually implementing the feature is on our to-do list: http://sourceware.org/bugzilla/show_bug.cgi?id=13588 Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: GDB - Can I See Every Symbol Called?
Michael == Michael Meeks michael.me...@suse.com writes: MichaelIIWY I'd insert a break-point thus: Michaelbreak unotxdoc.cxx:12345 Michaelwhere 12345 is the line-number of the lcl_ function that Michael is most likely inlined (as static 's tend to be) which usually Michael confuses gdb nicely ;-) [ when compiled with -g -O2 etc. ]. break inlinefunction should work ok with a recent-enough version of gdb. Older versions needed break file:line. Even older versions just didn't work at all. If you know of bugs here, please report them. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: GDB - Can I See Every Symbol Called?
Joel == Joel Madero jmadero@gmail.com writes: Joel I've finally gotten comfortable with gdb as far as break points and Joel what not. What I want to know is if there is a way for me to see all Joel symbols called during a given run of soffice. Getting a log that shows Joel every symbol called would allow me to compare to runs of libreoffice, Joel one where I push Ok the other where I push Cancel. It isn't really possible with gdb. I'd like it to be -- I wish rbreak * would DTRT -- but unfortunately some parts of gdb aren't as scalable as I'd like, and nobody has yet put the work into fixing them. I've personally never actually wanted to do this, but if I were going to tackle it I would either look at using ltrace, systemtap, or writing a valgrind skin (perhaps one exists already, I don't know). Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: gdb backtrace burpage ...
Michael == Michael Meeks michael.me...@suse.com writes: Michael #2 0xb784d6f1 in SfxFrame::Create (i_rFrame=warning: RTTI symbol not Michael found for class 'framework::Frame' Michael warning: RTTI symbol not found for class 'framework::Frame' Michael warning: RTTI symbol not found for class 'framework::Frame' Michael warning: RTTI symbol not found for class 'framework::Frame' Michael But is there any way of suppressing that warning thrash for the Michael case where it is not enabled ? There's no way to disable it. File a feature request in gdb bugzilla if you want that. I looked into this a little. The warning here is peculiar. Despite what it says, gdb is not actually looking for an RTTI symbol. Instead it is just looking for the debug info for the indicated type. I don't know a simple way to reproduce this problem, or I would dig into it a bit more. My guess is that it can only be reproduced in relatively odd situations, like where you have some subset of the debuginfo, but not quite all of it. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: gdb backtrace burpage ...
Christina == Christina Roßmanith chrrossman...@gmx.de writes: Christina You mean if I compile only a subset of modules with Christina dbglevel0? Maybe. I'm not really sure. I'd like to find a simple reproducer so I could find out. Christina I'm getting the RTTI messages as well but can't remember, Christina what I did immediately before enabling them...and I have Christina the feeling that gdb -tui crashes quite often since then. Please report all crashes to gdb bugzilla. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] concept for c++ based subsequenttests
Michael == Michael Meeks michael.me...@suse.com writes: Michael I guess we should add an 'easy hack' for some gdb / python Michael goodness to expand / annotate Basic stack frames prettily in Michael the debugger too ;-) [ we can but wish ]. FWIW Phil Muldoon is working on a new gdb/Python feature called frame filters that will let interpreters add additional data to backtraces. The idea is to be able to see the raw (implementation) stack frames and the filtered (interpreted) frames at the same time. This won't be in 7.4 but I hope it will be in the release after that. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] Even more debugging info
Lubos == Lubos Lunak l.lu...@suse.cz writes: Lubos As a sidenote, this gave me an interesting idea that I want to Lubos try somewhen. It might be actually helpful to explicitly not Lubos have debug info about macros and enclose bodies of some functions Lubos like uno::Reference::operator-() or boost::shared_ptr stuff in Lubos one huge macro that'd avoid stepping through them (which I really Lubos hate, as it's mostly pointless). CVS gdb has a 'skip' feature that can be used for this: (gdb) help skip Ignore a function while stepping. Usage: skip [FUNCTION NAME] If no function name is given, ignore the current function. [...] Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] Even more debugging info
Lionel == Lionel Elie Mamane lio...@mamane.lu writes: Tom In a recent-enough GCC (I don't know if it made 4.6, but anyway I Tom think it is in Fedora 16), there is a GNU extension to how macro Tom information is represented. This extension greatly reduces the size Tom of the macro information. Lionel And this GNU extension is enabled with -ggdb3? With -gdwarf-4? How Lionel do I see whether it is used? If you use -g3 and do not use -gstrict-dwarf, then the new macro style is used. You can check for the new style using: readelf -WS ./some-file | grep debug_macro The old section name is called .debug_macinfo. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] Even more debugging info
Stephan == Stephan Bergmann sberg...@redhat.com writes: Stephan On 11/30/2011 05:37 PM, Lionel Elie Mamane wrote: Any opinion about this patch? I have it in my local repo, and it helps me when running under gdb, as gdb now knows about macros! Stephan Would -ggdb3 excessively increase object size compared to -ggdb2? The short answer is yes, but there is a more complicated answer. First, one must consider the tradeoffs. I always use -g3 when building gdb, because gdb uses macros fairly heavily, and because I just don't care about the extra space. I prefer the convenience. YMMV of course, and I don't know enough about LibreOffice to offer an opinion. In a recent-enough GCC (I don't know if it made 4.6, but anyway I think it is in Fedora 16), there is a GNU extension to how macro information is represented. This extension greatly reduces the size of the macro information. If you really want to shrink debuginfo, use -gdwarf-4 and -fdebug-types-section. The former has been in GCC for a while, I don't remember when the latter was added. These options cause debuginfo for most big types to be shared across compilation units, a huge size win. This also makes gdb use less memory. The downside of this feature is that not all tools have been upgraded to understand it. gdb works fine, but the 7 dwarves, and systemtap, and I think valgrind, will barf. They will all come around eventually, though I am not sure when. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] gdb: catching exceptions...
Michael == Michael Stahl mstahl-h+wxahxf7alqt0dzr+a...@public.gmane.org writes: Mark Wielaard pointed out this note to me. Feel free to CC 'arc...@sourceware.org' on hacks like this, we're interested in hearing about them and to find out what developers need from gdb. Michael since my smoketest failed again today, i've distracted myself a bit Michael with GDB Python scripting, the result of which is a command that can Michael break when an exception is thrown, but only if a certain function Michael (which is given as parameter) is on the stack. Nice. Michael seems to work, but it is quite slow: finding the smoketest exception Michael in SfxBaseModel::getTitle takes GDB 4 minutes of CPU time. Michael now probably somebody will tell me that i'm too stupid to RTFM and Michael there is a much simpler way to do this :) There isn't a simpler way right now. You could try to do it by having a breakpoint on the function you care about that keeps a count of entries and exits and enables the __cxa_throw breakpoint if this is nonzero. However, this is hard to do reliably right now, because you have to do a lot of manual management of return breakpoints. I think this may get simpler when the finish breakpoint feature goes in: http://sourceware.org/ml/gdb-patches/2011-10/msg00394.html We're interested in hearing about real-life uses of exceptions and what better functionality gdb could provide. Freely file feature requests in gdb bugzilla; enhancing C++ development is a primary goal of ours these days. thanks, Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] gdb: catching exceptions...
I'm re-sending this note... I sent it yesterday via gmane, but I guess this list is configured for moderation even in that situation; so today I joined. Mark Wielaard pointed out this note to me. Feel free to CC 'arc...@sourceware.org' on hacks like this, we're interested in hearing about them and to find out what developers need from gdb. Michael since my smoketest failed again today, i've distracted myself a bit Michael with GDB Python scripting, the result of which is a command that can Michael break when an exception is thrown, but only if a certain function Michael (which is given as parameter) is on the stack. Nice. Michael seems to work, but it is quite slow: finding the smoketest exception Michael in SfxBaseModel::getTitle takes GDB 4 minutes of CPU time. Michael now probably somebody will tell me that i'm too stupid to RTFM and Michael there is a much simpler way to do this :) There isn't a simpler way right now. You could try to do it by having a breakpoint on the function you care about that keeps a count of entries and exits and enables the __cxa_throw breakpoint if this is nonzero. However, this is hard to do reliably right now, because you have to do a lot of manual management of return breakpoints. I think this may get simpler when the finish breakpoint feature goes in: http://sourceware.org/ml/gdb-patches/2011-10/msg00394.html We're interested in hearing about real-life uses of exceptions and what better functionality gdb could provide. Freely file feature requests in gdb bugzilla; enhancing C++ development is a primary goal of ours these days. thanks, Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] gdb: info mutex
Michael == Michael Stahl mstahl-h+wxahxf7alqt0dzr+a...@public.gmane.org writes: Michael have just found a great python command for GDB by Tom Tromey called Michael info mutex: Michael http://www.cygwin.com/ml/archer/2010-q3/msg00024.html Michael looks very helpful for debugging a deadlock: Just be warned, it is a bit broken. It has some heuristics to see when a thread is attempting to acquire a lock, but these do not always yield the correct answer. Still, it shouldn't be too misleading, as you can always dig around manually to verify what it says. Also I think you need glibc debuginfo installed for it to work at all. We hope to write a better version, replacing the heuristics with hidden breakpoints on sdt.h probe points in glibc. I think the probes might be there (in Fedora, but probably not anywhere else, as upstream rejected the probes), but we haven't written the gdb side yet. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] gdb: catching exceptions...
Third time's the charm. Apologies if you've seen this a lot. Mark Wielaard pointed out this note to me. Feel free to CC 'arc...@sourceware.org' on hacks like this, we're interested in hearing about them and to find out what developers need from gdb. Michael since my smoketest failed again today, i've distracted myself a bit Michael with GDB Python scripting, the result of which is a command that can Michael break when an exception is thrown, but only if a certain function Michael (which is given as parameter) is on the stack. Nice. Michael seems to work, but it is quite slow: finding the smoketest exception Michael in SfxBaseModel::getTitle takes GDB 4 minutes of CPU time. Michael now probably somebody will tell me that i'm too stupid to RTFM and Michael there is a much simpler way to do this :) There isn't a simpler way right now. You could try to do it by having a breakpoint on the function you care about that keeps a count of entries and exits and enables the __cxa_throw breakpoint if this is nonzero. However, this is hard to do reliably right now, because you have to do a lot of manual management of return breakpoints. I think this may get simpler when the finish breakpoint feature goes in: http://sourceware.org/ml/gdb-patches/2011-10/msg00394.html We're interested in hearing about real-life uses of exceptions and what better functionality gdb could provide. Freely file feature requests in gdb bugzilla; enhancing C++ development is a primary goal of ours these days. thanks, Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] [PATCH] fix xml2cmp build with svn gcc
Norbert == Norbert Thiebaud nthieb...@gmail.com writes: Norbert 1/ Are we sure that this behavior will be seen in a _released_ gcc ? Norbert (not a rhetorical question, I'm a bit weary of chasing gcc trunk that Norbert closely) I believe so; or at least, this is definitely a bug in LibreOffice. 'checkSize' is inherited from a dependent base class, so it must be qualified. See, e.g.: http://www.parashift.com/c++-faq-lite/templates.html#faq-35.19 Norbert 2/ I am not well equipped to apply in-lined patch like that (I use Norbert gmail as main mailbox for this dev-list... so it is cut-and-paste with Norbert horror stories about line wrapping and all) Norbert could you use git format-patch to generate patches and attach them Norbert (git format-patch as the added benefit that I don't have to type Norbert --author=Your Name you@email -m a nice commit message that _I_ Norbert would have to come up with but instead cat your_git_formated_patch Norbert | git am I will do that in the future, thanks for the tip. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] [PUSHED] fix xml2cmp build with svn gcc
Noel I rolled up this patch and the previous one here Noel http://cgit.freedesktop.org/libreoffice/core/commit/?id=c2a634416ba8f385c25c16cc52aeae6f68cab9db I don't understand where these commits go. I re-pulled in 'ure' (actually everywhere using ./g pull) but I don't see that commit. I'm on 'master', should I be using some other branch? Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] [PUSHED] fix xml2cmp build with svn gcc
Miklos Yes, core.git. See Miklos http://wiki.documentfoundation.org/Development/Native_Build#Getting_the_sources Miklos Background: http://wiki.documentfoundation.org/Development/One_Git_Conversion Thanks. I must have read some stale document before starting; but unfortunately I don't recall what. I will start over. Tom ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
[Libreoffice] [PATCH] fix soltools build with svn gcc
I tried building LibreOffice with svn trunk gcc. It dies in soltools with: Compiling: soltools/giparser/gi_list.cxx In file included from ../inc/gi_list.hxx:33:0, from /home/tromey/Space/LibreOffice/bootstrap/soltools/giparser/gi_list.cxx:33: ../inc/st_list.hxx: In instantiation of 'DynamicListXY DynamicListXY::operator=(const DynamicListXY) [with XY = GenericInfo]': /home/tromey/Space/LibreOffice/bootstrap/soltools/giparser/gi_list.cxx:59:25: required from here ../inc/st_list.hxx:298:10: error: 'push_back' was not declared in this scope, and no declarations were found by argument-dependent lookup at the point of instantiation [-fpermissive] ../inc/st_list.hxx:298:10: note: declarations in dependent base 'ST_ListGenericInfo*' are not found by unqualified lookup ../inc/st_list.hxx:298:10: note: use 'this-push_back' instead ../inc/st_list.hxx: In instantiation of 'void DynamicListXY::Insert(unsigned int, XY* const) [with XY = GenericInfo]': /home/tromey/Space/LibreOffice/bootstrap/soltools/giparser/gi_list.cxx:234:1: required from here ../inc/st_list.hxx:311:5: error: 'checkSize' was not declared in this scope, and no declarations were found by argument-dependent lookup at the point of instantiation [-fpermissive] ../inc/st_list.hxx:311:5: note: declarations in dependent base 'ST_ListGenericInfo*' are not found by unqualified lookup ../inc/st_list.hxx:311:5: note: use 'this-checkSize' instead The appended patch fixes this problem by adding 'this-' qualifiers. This is contributed under the LGPLv3+/MPL. Tom diff --git a/soltools/inc/st_list.hxx b/soltools/inc/st_list.hxx index 51f0a32..69b6943 100644 --- a/soltools/inc/st_list.hxx +++ b/soltools/inc/st_list.hxx @@ -295,7 +295,7 @@ DynamicListXY::operator=( const DynamicListXY i_rList ) it != i_rList.end(); ++it ) { - push_back( new XY(*(*it)) ); + this-push_back( new XY(*(*it)) ); } return *this; } @@ -308,7 +308,7 @@ DynamicListXY::Insert(unsigned pos, XY * const elem_) if ( pos this-len ) return; -checkSize(DynamicListXY::len+2); +this-checkSize(DynamicListXY::len+2); memmove( DynamicListXY::inhalt+pos+1, DynamicListXY::inhalt+pos, (DynamicListXY::len-pos) * sizeof(XY*) ); this-inhalt[pos] = elem_; this-len++; ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
[Libreoffice] [PATCH] fix xml2cmp build with svn gcc
I tried building LibreOffice with svn trunk gcc. It dies in xml2cmp with: Compiling: xml2cmp/source/xcd/xmlelem.cxx In file included from ./xmlelem.hxx:38:0, from /home/tromey/Space/LibreOffice/bootstrap/clone/ure/xml2cmp/source/xcd/xmlelem.cxx:30: ./../support/list.hxx: In instantiation of 'void DynamicListXY::insert(unsigned int, XY* const) [with XY = XmlElement]': /home/tromey/Space/LibreOffice/bootstrap/clone/ure/xml2cmp/source/xcd/xmlelem.cxx:263:1: required from here ./../support/list.hxx:229:5: error: 'checkSize' was not declared in this scope, and no declarations were found by argument-dependent lookup at the point of instantiation [-fpermissive] ./../support/list.hxx:229:5: note: declarations in dependent base 'ListXmlElement*' are not found by unqualified lookup ./../support/list.hxx:229:5: note: use 'this-checkSize' instead dmake: Error code 1, while making '../../unxlngx6.pro/obj/xmlelem.obj' The appended patch fixes this problem by adding a 'this-' qualifier. This is contributed under the LGPLv3+/MPL. Tom diff --git a/xml2cmp/source/support/list.hxx b/xml2cmp/source/support/list.hxx index 6de123e..e4e5361 100644 --- a/xml2cmp/source/support/list.hxx +++ b/xml2cmp/source/support/list.hxx @@ -226,7 +226,7 @@ DynamicListXY::insert(unsigned pos, XY * const elem_) if ( pos this-len ) return; -checkSize(this-len+2); +this-checkSize(this-len+2); memmove(this-inhalt[pos+1], this-inhalt[pos], (this-len-pos) * sizeof(XY*) ); this-inhalt[pos] = elem_; this-len++; ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
[Libreoffice] [PATCH] fix cosv build
I tried building LibreOffice with svn trunk gcc; though in this case the bug seems to be a fairly generic include order problem. The build dies in cosv with: Compiling: cosv/source/strings/string.cxx In file included from ../../inc/cosv/string.hxx:33:0, from ../../inc/cosv/csv_precomp.h:39, from ../inc/precomp.h:32, from /home/tromey/Space/LibreOffice/bootstrap/clone/sdk/cosv/source/strings/string.cxx:29: ../../inc/cosv/stringdata.hxx: In instantiation of 'csv::StringDataCHAR::StringData(const CHAR*, csv::StringDataCHAR::size_type) [with CHAR = char, csv::StringDataCHAR::size_type = long unsigned int]': /home/tromey/Space/LibreOffice/bootstrap/clone/sdk/cosv/source/strings/string.cxx:75:17: required from here ../../inc/cosv/stringdata.hxx:105:5: error: 'memcpy' was not declared in this scope, and no declarations were found by argument-dependent lookup at the point of instantiation [-fpermissive] /usr/include/string.h:44:14: note: 'void* memcpy(void*, const void*, size_t)' declared here, later in the translation unit dmake: Error code 1, while making '../../unxlngx6.pro/obj/string.obj' string.hxx includes string.h, which declares memcpy, after cosv/stringdata.hxx. Moving string.h earlier fixes the problem. This is contributed under the LGPLv3+/MPL. Tom diff --git a/cosv/inc/cosv/string.hxx b/cosv/inc/cosv/string.hxx index 08b1220..8f249f9 100644 --- a/cosv/inc/cosv/string.hxx +++ b/cosv/inc/cosv/string.hxx @@ -30,9 +30,9 @@ #define COSV_STRING_HXX // USED SERVICES +#include string.h #include cosv/stringdata.hxx #include cosv/str_types.hxx -#include string.h #include cosv/csv_ostream.hxx #include vector ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
[Libreoffice] [PATCH] sax build -vs- svn gcc
I tried building LibreOffice with svn trunk gcc. It dies in sax with: In file included from /home/tromey/Space/LibreOffice/bootstrap/sax/source/tools/fastserializer.cxx:33:0: /home/tromey/Space/LibreOffice/bootstrap/solver/350/unxlngx6.pro/inc/comphelper/sequenceasvector.hxx: In instantiation of ‘void comphelper::SequenceAsVectorTElementType::operator(const com::sun::star::uno::SequenceT) [with TElementType = int]’: /home/tromey/Space/LibreOffice/bootstrap/solver/350/unxlngx6.pro/inc/comphelper/sequenceasvector.hxx:111:13: required from ‘comphelper::SequenceAsVectorTElementType::SequenceAsVector(const com::sun::star::uno::SequenceT) [with TElementType = int]’ /home/tromey/Space/LibreOffice/bootstrap/sax/source/tools/fastserializer.cxx:468:55: required from here /home/tromey/Space/LibreOffice/bootstrap/solver/350/unxlngx6.pro/inc/comphelper/sequenceasvector.hxx:150:17: error: ‘push_back’ was not declared in this scope, and no declarations were found by argument-dependent lookup at the point of instantiation [-fpermissive] /home/tromey/Space/LibreOffice/bootstrap/solver/350/unxlngx6.pro/inc/comphelper/sequenceasvector.hxx:150:17: note: declarations in dependent base ‘std::vectorint, std::allocatorint ’ are not found by unqualified lookup /home/tromey/Space/LibreOffice/bootstrap/solver/350/unxlngx6.pro/inc/comphelper/sequenceasvector.hxx:150:17: note: use ‘this-push_back’ instead The appended patch fixes this problem by adding a 'this-' qualifier. This is contributed under the LGPLv3+/MPL. Tom diff --git a/o3tl/inc/o3tl/vector_pool.hxx b/o3tl/inc/o3tl/vector_pool.hxx index 6ef4e96..28299f0 100644 --- a/o3tl/inc/o3tl/vector_pool.hxx +++ b/o3tl/inc/o3tl/vector_pool.hxx @@ -65,7 +65,7 @@ namespace o3tl } else { -push_back(value_type(rCopy)); +this-push_back(value_type(rCopy)); return this-size()-1; } } ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] [PATCH] sax build -vs- svn gcc
Tom == Tom Tromey tro...@redhat.com writes: Tom I tried building LibreOffice with svn trunk gcc. Tom It dies in sax with: [...] Tom The appended patch fixes this problem by adding a 'this-' qualifier. Oops, I appended the wrong patch to this message. Here is the correct patch. I am not really sure where this file comes from. From what I can tell it isn't in git. Tom --- sequenceasvector.hxx.~1~2011-08-10 08:25:19.179720480 -0600 +++ sequenceasvector.hxx2011-08-10 11:48:05.404703517 -0600 @@ -147,7 +147,7 @@ const TElementType* pSource = lSource.getConstArray(); for (sal_Int32 i=0; ic; ++i) -push_back(pSource[i]); +this-push_back(pSource[i]); } //--- ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice