On 24/01/2014 02:03, Charles Gordon wrote: > On 23 janv. 2014, at 14:40, François Revol <re...@free.fr> wrote: > >> On 23/01/2014 14:02, Charles Gordon wrote: >>> Hi Francois, >>> >>> I just committed a large patch to fix this issue among other malloc related >>> inconsistencies. >>> valgrind was right, it was a bug! >>> qe_mallocz would have cleared the allocated structure, but you used malloc >>> in haiku.cpp to allocate it :( >>> valgrind rules! >> >> Actually I ran it on GNU/Linux since valgrind isn't available yet on >> Haiku... (I just started the port). >> The X11 code did use qe_mallocz, but I didn't check the tty code though. > > Then this issue is unresolved. > I fixed most memory leaks using valgrind, but I cannot test x11 on Linux for > the moment.
It seems to be fixed here. I believe it's the qe_malloc in tty.c at font creation that you changed to qe_mallocz. On Linux both X11 and tty support are built, so it probably initialized the tty display regardless. As for the --free-all option, I don't see the need, if you want to free at exit then do it, but there is no need for an option :p It makes a difference: ==450== in use at exit: 934,064 bytes in 894 blocks vs. ==451== in use at exit: 71,025 bytes in 497 blocks But it frees a little too much maybe: ==451== Invalid read of size 8 ==451== at 0x41B104: main (qe.c:8111) ==451== Address 0x647b1f0 is 0 bytes inside a block of size 24 free'd ==451== at 0x4C2A70C: free (vg_replace_malloc.c:468) ==451== by 0x41B12B: main (qe.c:8112) So I'd say just don't bother. François. _______________________________________________ Qemacs-devel mailing list Qemacs-devel@nongnu.org https://lists.nongnu.org/mailman/listinfo/qemacs-devel
