* setting breakpoints with C++ typedef names [was: Re: debug information] [not found] ` <200611212225.28377.pogonyshev@gmx.net> @ 2006-11-22 12:03 ` Benjamin Kosnik 2006-11-22 13:35 ` Daniel Jacobowitz 2006-11-24 21:11 ` Martin Sebor 0 siblings, 2 replies; 4+ messages in thread From: Benjamin Kosnik @ 2006-11-22 12:03 UTC (permalink / raw) To: Paul Pogonyshev; +Cc: libstdc++, gdb > > But still function names include full type name: `b 'bar(basic_pod<int>)''. > Not much of a problem in this case, but sometimes the full name is very > long... Oh. I finally understand you. Yes, this is a drag. You want to be able to do: (gdb) b bar(ipod) instead of (gdb) b bar(basic_pod<int>) As you've noticed, this doesn't currently work. (x86-linux, mainline gcc, gdb (6.5-13.fc6rh)) (gdb) b bar(ipod) Function "bar(ipod)" not defined. As you stated, you have to: (gdb) b bar(basic_pod<int>) Breakpoint 4 at 0x80489d7: file debug.cc, line 24. I suggest reporting this to the gdb list (gdb@sourceware.org). You might want to title that email something like "setting breakpoints with C++ typedef names" or something more descriptive. Here, I've done that for you. To me, it looks like enough information is present in the debuginfo to support this kind of usage. However, I am no expert in this area. -benjamin ps. Here's the test case template<typename T> struct basic_pod { typedef T value_type; value_type _M_data; }; typedef basic_pod<int> ipod; void bar(ipod p) { p._M_data = 0; } void bar(int i) { ++i; } int main() { ipod p; bar(p); // b bar(ipod) bar(5); // b bar(int) return 0; } ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: setting breakpoints with C++ typedef names [was: Re: debug information] 2006-11-22 12:03 ` setting breakpoints with C++ typedef names [was: Re: debug information] Benjamin Kosnik @ 2006-11-22 13:35 ` Daniel Jacobowitz 2006-11-24 21:11 ` Martin Sebor 1 sibling, 0 replies; 4+ messages in thread From: Daniel Jacobowitz @ 2006-11-22 13:35 UTC (permalink / raw) To: Benjamin Kosnik; +Cc: Paul Pogonyshev, libstdc++, gdb On Wed, Nov 22, 2006 at 01:03:41PM +0100, Benjamin Kosnik wrote: > I suggest reporting this to the gdb list (gdb@sourceware.org). You > might want to title that email something like "setting breakpoints with > C++ typedef names" or something more descriptive. Here, I've done that > for you. > > To me, it looks like enough information is present in the debuginfo to > support this kind of usage. However, I am no expert in this area. I don't think that GDB will ever support this. It's one of a batch of C++ improvements that would be nice to have, but getting it right would require most of a C++ front end. And for various reasons, I don't think that reusing the g++ front end would be practical. We could probably get it to work in strictly limited cases, though. -- Daniel Jacobowitz CodeSourcery ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: setting breakpoints with C++ typedef names [was: Re: debug information] 2006-11-22 12:03 ` setting breakpoints with C++ typedef names [was: Re: debug information] Benjamin Kosnik 2006-11-22 13:35 ` Daniel Jacobowitz @ 2006-11-24 21:11 ` Martin Sebor 2006-11-24 21:28 ` Paul Pogonyshev 1 sibling, 1 reply; 4+ messages in thread From: Martin Sebor @ 2006-11-24 21:11 UTC (permalink / raw) To: Benjamin Kosnik; +Cc: Paul Pogonyshev, libstdc++, gdb Benjamin Kosnik wrote: >>But still function names include full type name: `b 'bar(basic_pod<int>)''. >>Not much of a problem in this case, but sometimes the full name is very >>long... > > > Oh. I finally understand you. Yes, this is a drag. > > You want to be able to do: > > (gdb) b bar(ipod) > > instead of > > (gdb) b bar(basic_pod<int>) This is especially troublesome when the real type is private to the implementation, as in, for example, void foo (std::map<T, U>::iterator); In order for a user to set a breakpoint on foo they need to know that the type of libstdc++'s map<T, U>::iterator is really called std::_Rb_tree_iterator<std::pair<const T, U> > (and something completely different when using a different implementation of the library). I expect this use of private names to become even more pervasive when (and if) typedef templates are implemented. Martin ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: setting breakpoints with C++ typedef names [was: Re: debug information] 2006-11-24 21:11 ` Martin Sebor @ 2006-11-24 21:28 ` Paul Pogonyshev 0 siblings, 0 replies; 4+ messages in thread From: Paul Pogonyshev @ 2006-11-24 21:28 UTC (permalink / raw) To: Martin Sebor; +Cc: Benjamin Kosnik, libstdc++, gdb Martin Sebor wrote: > Benjamin Kosnik wrote: > >>But still function names include full type name: `b 'bar(basic_pod<int>)''. > >>Not much of a problem in this case, but sometimes the full name is very > >>long... > > > > > > Oh. I finally understand you. Yes, this is a drag. > > > > You want to be able to do: > > > > (gdb) b bar(ipod) > > > > instead of > > > > (gdb) b bar(basic_pod<int>) > > This is especially troublesome when the real type is private to > the implementation, as in, for example, > > void foo (std::map<T, U>::iterator); > > In order for a user to set a breakpoint on foo they need to know > that the type of libstdc++'s map<T, U>::iterator is really called > std::_Rb_tree_iterator<std::pair<const T, U> > > (and something completely different when using a different > implementation of the library). Another important problem (and most important to me, since I set up breakpoints using Tab-completion anyway) is that it makes backtraces and similar stuff virtually unreadable. Paul ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2006-11-24 21:28 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <200611191342.48724.pogonyshev@gmx.net>
[not found] ` <20061121100701.6b980f48.bkoz@redhat.com>
[not found] ` <200611212225.28377.pogonyshev@gmx.net>
2006-11-22 12:03 ` setting breakpoints with C++ typedef names [was: Re: debug information] Benjamin Kosnik
2006-11-22 13:35 ` Daniel Jacobowitz
2006-11-24 21:11 ` Martin Sebor
2006-11-24 21:28 ` Paul Pogonyshev
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox