Hello,
both bugs 1038594 and 1196143 appear to be 'fixed' by not g_free'ing gr->color. This is a text gradient buffer mangled in both bugs (in 1196143 you can see some destruction of the picture when repainting after dialog release). In 1038594, valgrind tells me about attempted reads from the same address which was freed before, and gtk likely reuses the memory, and this fast.
This would mean that there is a copy of the color buf somewhere that is not updated but I still am not well into the code.
I would propose memory handling of this buffer to be moved into GC, before the release, or using malloc/free if that works.
https://sourceforge.net/tracker/?func=detail&atid=604306&aid=1038594... https://sourceforge.net/tracker/?func=detail&atid=604306&aid=1196143...
Regards, ralf