On Sat, 17 Mar 2001, gerard patel wrote:
First thanks for the bug report.
The trace is not difficult to read, the menu item is a text item and the text pointer is NULL. MENU_CalcItemSize is not handling this case. What is strange is this problem appears only with my patch. Maybe
Now hang on! I didn't say that. I said I had first provoked it after applying your patch. :-). I guess I should have reverted the patch so I would know for sure before I said anything, but I figured you had just worked in that area and might have a better idea about fixing it than I just jumping in and putting a test for null or a __try around it.
I don't use GUI very much, and that crash could have been waiting for years to be provoked.
I will revert presently and repeat the test in case that might help.
your app is really initializing the menu when receiving the WM_INITMENU message, but *only* when there is need to display the menu. I can't see how it could find out about it, though.
Anyway, could you try the following patch ? It tries to remove the main case I saw where the app can set a menu item to text type with a null text pointer. It is a legitimate patch by itself, even if it does not fix your problem. Maybe it could have nasty side effects. If it avoids the systematic crash, try to find if the other is really gone for good too.
Still crashes. The patch is in the source, the object file and libuser32.so are newer than menu.c, so I _think_ there must be another way to get a null text pointer with a text type.
This time I spelled "info local" right. tee is a wonderful little program.
I have found a strangeness in the HEAP_strdup macros - WARNING - this is of interest to anyone dabbling with Wine code. While HEAP_strdupW and HEAP_strdupAtoW appear very similar, the specs are not the same; one returns a null pointer when called with a null pointer, the other allocates 2 bytes of memory and returns a pointer to it. Nice trap :-)
Gerard
Lawson