segfaults in gnc-html-guppi.c

Christian Stimming stimming@tuhh.de
Sat, 25 Aug 2001 01:20:27 +0200


-----BEGIN PGP SIGNED MESSAGE-----

I was debugging the [xy]-axis labels of the Guppi barcharts. While I tried 
those, I encountered mysterious segfaults caused from gnc-html-guppi.c. Those 
seemed to appear after too many subsequent calls to g_hash_table_lookup with 
negative result. GDB said this:

Program received signal SIGSEGV, Segmentation fault.
#0  0x400eb10b in g_hash_table_lookup () from /usr/lib/libglib-1.2.so.0
#1  0x80d21b7 in gnc_html_embedded_barchart (parent=0x1, w=400, h=400,
    params=0x4) at gnc-html-guppi.c:784
#2  0x8132a46 in gnc_account_tree_info.118 ()
Cannot access memory at address 0x4

To reproduce: First, get a Guppi CVS version newer than 05/25/2001 and, yes, 
the current Guppi CVS is broken, but everything up to two weeks ago will work 
fine. Second, change the #define in current CVS's gnc-html-guppi.c:606 from 0 
(false) to 1 (true). Then start gnucash and open the Asset Barchart report. 
Crash.

Note the GDB message's line number gnc-html-guppi.c:784 (or as well 785): 
This is the 19th call (of 20) of g_hash_table_lookup in a row. If the #define 
is set to false, you will have 16 of them, and if your Guppi is 0.35.5 or 
small you will still have a maximum of 18 of those calls. Hence, this error 
won't occur unless you have things set up in the described way.

Can anybody reproduce that? Can anyone get more information out of his 
debugger?

Christian
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.5 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iQCVAwUBO4bhQWXAi+BfhivFAQF7GAQAiNM2QH995C5ha7+arGfqVZ4f0jlCuZmJ
nNrWTXDP3cPNoqqymAtZtVvZDftTzwEa65Tqxz+XK2/3Vq5AWbtZmhFT0WLErDH6
HFTIelp/1R2z7yXzwwo0NJhrM+x8DrDD2PcTiNR4xaqBDMjGXrApHOens+zC+21x
DLSE2jh55nA=
=BA1x
-----END PGP SIGNATURE-----