Dominique Pelle wrote:
> On Sun, Jun 8, 2008 at 2:25 PM, Bram Moolenaar <[EMAIL PROTECTED]> wrote: > > > Dominique Pelle wrote: > > > >> On Sat, Jun 7, 2008 at 8:44 AM, Dominique Pelle wrote: > >> > >> > Attached: > >> > - crash-os_unix.c.patch > >> > - leak-os_unix.c.patch > >> > >> I built the gui-motif, gui-athena and gui-gtk2 on Linux x86 and the crash > >> in > >> os_unix.c only happens with gui-motif (I'm using LessTif). > >> > >> So instead of doing #if 0 as in my previous patch, it should be better to > >> do #ifndef FEAT_GUI_MOTIF. > >> > >> gui-athena shows errors with -DEXITFREE just a few lines below though: > >> > >> ==15738== Invalid read of size 4 > >> ==15738== at 0x42CDEE7: XCloseDisplay (in /usr/lib/libX11.so.6.2.0) > >> ==15738== by 0x8140419: mch_free_mem (os_unix.c:2900) > >> ==15738== by 0x810A969: free_all_mem (misc2.c:1125) > >> ==15738== by 0x8140573: mch_exit (os_unix.c:3012) > >> ==15738== by 0x80DE4C8: getout (main.c:1342) > >> ==15738== by 0x80A56C3: ex_quit_all (ex_docmd.c:6281) > >> ==15738== by 0x809F8CB: do_one_cmd (ex_docmd.c:2623) > >> ==15738== by 0x809D18C: do_cmdline (ex_docmd.c:1099) > >> ==15738== by 0x812030E: nv_colon (normal.c:5179) > >> ==15738== by 0x8119B88: normal_cmd (normal.c:1152) > >> ==15738== by 0x80DE20A: main_loop (main.c:1181) > >> ==15738== by 0x80DDD5A: main (main.c:940) > >> ==15738== Address 0x44020cc is 148 bytes inside a block of size 1,340 > >> free'd > >> ==15738== at 0x402265C: free (vg_replace_malloc.c:323) > >> ==15738== by 0x42E0F3E: _XFreeDisplayStructure (in > >> /usr/lib/libX11.so.6.2.0) > >> ==15738== by 0x42CDFC5: XCloseDisplay (in /usr/lib/libX11.so.6.2.0) > >> ==15738== by 0x424EC49: (within /usr/lib/libXt.so.6.0.0) > >> ==15738== by 0x424EE36: XtCloseDisplay (in /usr/lib/libXt.so.6.0.0) > >> ==15738== by 0x424EE70: (within /usr/lib/libXt.so.6.0.0) > >> ==15738== by 0x424F164: XtDestroyApplicationContext (in > >> /usr/lib/libXt.so.6.0.0) > >> ==15738== by 0x81403F4: mch_free_mem (os_unix.c:2896) > >> ==15738== by 0x810A969: free_all_mem (misc2.c:1125) > >> ==15738== by 0x8140573: mch_exit (os_unix.c:3012) > >> ==15738== by 0x80DE4C8: getout (main.c:1342) > >> ==15738== by 0x80A56C3: ex_quit_all (ex_docmd.c:6281) > >> ==15738== by 0x809F8CB: do_one_cmd (ex_docmd.c:2623) > >> ==15738== by 0x809D18C: do_cmdline (ex_docmd.c:1099) > >> ==15738== by 0x812030E: nv_colon (normal.c:5179) > >> ==15738== by 0x8119B88: normal_cmd (normal.c:1152) > >> ==15738== by 0x80DE20A: main_loop (main.c:1181) > >> ==15738== by 0x80DDD5A: main (main.c:940) > >> > >> os_unix.c: > >> > >> 2890 # if (defined(FEAT_X11) && defined(FEAT_XCLIPBOARD)) || > >> defined(PROTO) > >> 2891 if (xterm_Shell != (Widget)0) > >> 2892 XtDestroyWidget(xterm_Shell); > >> 2893 if (xterm_dpy != NULL) > >> 2894 XtCloseDisplay(xterm_dpy); > >> 2895 if (app_context != (XtAppContext)NULL) > >> !! 2896 XtDestroyApplicationContext(app_context); > >> 2897 # endif > >> 2898 # ifdef FEAT_X11 > >> 2899 if (x11_display != NULL && x11_display != xterm_dpy) > >> !! 2900 XCloseDisplay(x11_display); > >> 2901 # endif > >> > >> Line 2900 is using memory which was freed at line 2896 > >> according to valgrind. So it seems that > >> XtDestroyApplicationContext(app_context) is also > >> destroying x11_display. > >> > >> Attached patch fixes the error. > > > > Thanks. I wonder if we can use #ifndef LESSTIF_VERSION instead of > > FEAT_GUI_MOTIF. The other "#if 0" suggests that Solaris also has this > > problem. But I'm not sure about this specific code, since it was added > > much later. > > > Yes, #ifndef LESSTIF_VERSION is better than #ifndef FEAT_GUI_MOTIF. > I tested it with both lesstif and Open Motif, it's fine. After the attached > patch, neither lesstif, nor Open Motif access freed memory. > > Open Motif needs the same fix as for athena gui i.e. set x11_display > to NULL after 'XtDestroyApplicationContext(app_context);' since it > seems to take care of already calling 'XCloseDisplay(x11_display)'. > > Lesstif and Open Motif also use uninitialized memory but I assume > that's a problem in those libs and not with Vim. > > There are several things not freed with all GUIs (athena, open motif > lesstif, gtk2, gnome2) when exiting even with -DEXITFREE but that's > not a major problem I think. I fixed at least the following leak which > happens at least for Motif and athena gui (patch: leak-gui_x11.c). > > ==23613== 32 bytes in 1 blocks are definitely lost in loss record 34 of 126 > ==23613== at 0x4022AB8: malloc (vg_replace_malloc.c:207) > ==23613== by 0x40A8C08: xpmParseDataAndCreate (in /usr/lib/libXm.so.2.0.1) > ==23613== by 0x40AA0A6: XpmCreateImageFromData (in /usr/lib/libXm.so.2.0.1) > ==23613== by 0x40AA342: XpmCreatePixmapFromData (in > /usr/lib/libXm.so.2.0.1) > ==23613== by 0x81C9E0E: gui_mch_init (gui_x11.c:1541) > ==23613== by 0x81BA6F1: gui_init (gui.c:457) > ==23613== by 0x81A56BD: set_termname (term.c:1845) > ==23613== by 0x81A609D: termcapinit (term.c:2514) > ==23613== by 0x81BA0BE: gui_start (gui.c:91) > ==23613== by 0x80DED45: main (main.c:634) Yes, it's tough fixing all memory leaks for GUIs. The documentation isn't always clear about when memory can be freed. I'll include the ones the patched, thanks. -- hundred-and-one symptoms of being an internet addict: 14. You start introducing yourself as "Jim at I-I-Net dot net dot au" /// Bram Moolenaar -- [EMAIL PROTECTED] -- http://www.Moolenaar.net \\\ /// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\ \\\ download, build and distribute -- http://www.A-A-P.org /// \\\ help me help AIDS victims -- http://ICCF-Holland.org /// --~--~---------~--~----~------------~-------~--~----~ You received this message from the "vim_dev" maillist. For more information, visit http://www.vim.org/maillist.php -~----------~----~----~----~------~----~------~--~---
