Menu
▾
▴
#4811 NIPPON: Assertion at Dough's House (regression)
Features compiled in: Vorbis FLAC MP3 RGB zLib
When Dough is holding the letter from the mailbox to read it i get an assertion "label < label.size()" failed: file "engines/parallaction/graphics.cpp", line 565
No save position due to bug item #2969234
Nippon Safes Inc. (Multi-lingual/Amiga)
Amiga OS4
gcc (GCC) 4.2.4 (adtools build 20090118)
Open Mailbox, take letter, crash :-)
The very same happens when Donna is holding the Suzy Yong poster inside "Lucky Look"
Walk inside "Lucky Look" and wave around the poster
The label rendered for the text "examine the poster"' is freed when entering "Lucky Look" (in changeLocation, even two or three times). In fact all inventory labels are destroyed when changing locations, but it only triggers asserts when you have more inventory items than there are labels in the new room. (Since the label IDs are restarted at 0.)
This is another regression, which was specifically caused by revision 32873.
A rough solution to the problem
I attached a patch to fix the issue, by letting the zones to whom the label belongs to manage their lifecycle.
I tried both the attached savegames and I got no crashes, so at least the bug is patched. What I could not do is checking for memory leaks, that are likely to be.
Nice work. This fixes the crashes for me too.
Looking over the patch, I think it would leak the ad-hoc created labels for things like the test result screen. (Or at least I do not see where the labels created there are freed.)
A quick valgrind session shows label leaks on the language screen.
What is the status of this item?
I committed an enhanced version of the patch on r48811. I could not perform any leak checking with valgrind, and won't be able until I go back home in 10 days, so I did not backport it to the release branch.
This bug has been fixed in revision 48811 and 48827. The patch has been backported to the release branch of version 1.1.1 in r48828. The leaks have been plugged, except for the labels allocated by _c_testResults, which is invoked at most once per run.
Should this be closed or just priority lowered?
r48837 fixes the memory leaks.