Bogus warning/internal error with psymtabs/symtabs between gdb-6.8 and gcc-4.3.x

[Bernd Strieder]

Hello,

not sure whether there is a bug in gdb, or whether something 
between gcc and gdb is out of sync. My first guess is there is a 
problem in gdb. I have never observed this with gcc releases 
before 4.3 or gdb releases before 6.8 myself. The debugged 
application is some computer algebra stuff with a GUI ontop. The 
exe is 112MB large with debugging symbols.

I'm getting when compiled with g++-4.3.1 prerelease as delivered 
with openSUSE 11.0 or a vanilla g++-4.3.2 the following messages 
about internal errors in gdb, e.g. when using the command bt.

(gdb) bt
#0  Xs_StandardStringCompletion::doCompletion (this=0x8ac5888)
    at /home/strieder/xssr/sys/StandardStringCompletion.cc:1138
warning: (Internal error: pc 0x86b0906 in read in psymtab, but not 
in symtab.)

#1  0x086b0907 in XSSRMainWnd::CompletionThread::run (this=warning: 
(Internal error: pc 0x86b0906 in read in psymtab, but not in 
symtab.)

warning: (Internal error: pc 0x86b08b2 in read in psymtab, but not 
in symtab.)

warning: (Internal error: pc 0x86b0906 in read in psymtab, but not 
in symtab.)

0x8a7e768)
    at /home/strieder/xssr/sys/xssr-gui/XSSRMainWnd.cpp:442
#2  0xb7a0624f in QThreadInstance::start () 
from /usr/lib/libqt-mt.so.3
#3  0xb77be175 in start_thread () from /lib/libpthread.so.0
#4  0xb761ddce in clone () from /lib/libc.so.6


The fun part is, that those warnings will not occurr, after I used 
the command 'maint check-symtabs' and scroll far enough through 
the list by holding down return. As far as I understood during 
check-symtabs the partial symbol tables might be brought into sync 
with the real symbols, or whatever happens. I'm observing that gdb 
has a steady increase in memory utilization during check-symtabs.

Basically in check-symtabs all symbols of my app are listed like

Global symbol 
`_ZSt22__uninitialized_copy_aIPPN4Util14CISSSuffixTree9InnernodeINS1_12Block8TraitsIN9__gnu_cxx17__normal_iteratorIPK9Xs_LetterSt6vectorIS6_SaIS6_EEEEN41Xs_WeightedCombinedStringEquationSetForKB14ImplementationINSD_30UnselectedEquationDataImplicitEE30EarlyCPEliminatorConstIteratorEEEEESL_SK_ET0_T_SN_SM_RSaIT1_E'
 
only found 
in /home/strieder/xssr/sys/WeightedCombinedStringEquationSetForKB.cc 
psymtab

Since gdb seems to work in all other respects there is only that 
bogus warning/internal error message I'm asking about. I have not 
yet tried some recent snapshot of gdb, if there is a chance that 
the problem might have been fixed in one, I will give it a try. I 
think it might be difficult to create a small testcase (I tried), 
because for small symtabs gdb does not seem to use any caching 
mechanisms and just loads them at once.

The problem seems to have occurred earlier in similar ways, which I 
found in the net. This is what made me trying that maint 
check-symtabs. I have no further ideas on this. Please CC me, I 
have not subscribed for the list for just this issue, but I will 
look at the thread in the archives, anyway.

Kind regards,

Bernd Strieder