GNOME Bugzilla – Bug 100008
Leak when reading configuration data?
Last modified: 2004-12-22 21:47:04 UTC
This trace looks to be a leak in gnome-vfs. ==26956== 17 bytes in 2 blocks are definitely lost in loss record 24 of 93 ==26956== at 0x40167140: malloc (vg_clientfuncs.c:100) ==26956== by 0x41004959: g_malloc (gmem.c:137) ==26956== by 0x4097C731: module_path_element_new (gnome-vfs-configuration.c:80) ==26956== by 0x4097CBE2: parse_line (gnome-vfs-configuration.c:295) ==26956== by 0x4097CCCA: parse_file (gnome-vfs-configuration.c:338) ==26956== by 0x4097CEA6: configuration_load (gnome-vfs-configuration.c:387) ==26956== by 0x4097D165: gnome_vfs_configuration_init (gnome-vfs-configuration.c:475) ==26956== by 0x4097FADE: gnome_vfs_init (gnome-vfs-init.c:132) ==26956== by 0x408EBAB1: ??? (gnome-init.c:409) ==26956== by 0x408E85BD: gnome_program_postinit (gnome-program.c:1626) ==26956== by 0x408E89CD: gnome_program_initv (gnome-program.c:1873) ==26956== by 0x408E864B: gnome_program_init (gnome-program.c:1679) ==26956== by 0x805DF4F: main (gedit2.c:332) ==26956== by 0x41066656: __libc_start_main (../sysdeps/generic/libc-start.c:129) ==26956== by 0x805C920: (within /gnome/head/INSTALL/bin/gedit)
I think this is caused by gnome-vfs-configuration.c:293: g_hash_table_insert (configuration->method_to_module_path, method_name, element); I think this might be called more then once with the same method_name (really only guessing here). If that is the case, suspect that that's not the way it should be. Anyway if it's OK that entry in the hash table is replaced (or it is cause by something else) I made a patch that fixes the problem by using a hashtable with a value_destroy_func instead so even replacing doesn't produce memleaks. I really don't know the code, so be careful with my conclusions
Created attachment 12661 [details] [review] patch that fixes the memleak, possibly obscuring a real problem
See a different patch at bug #100682
*** Bug 100682 has been marked as a duplicate of this bug. ***
I think you are still leaking the key (method name) name when you try to insert a value for an existing method name (see line 296 of gnome-vfs-configuration.c).
Should the pri on this be higher due to the patch?
I committed the attached patch
Created attachment 14225 [details] [review] committed patch (I added a g_free in the g_hash_table_new_full)