Date: Fri, 15 Dec 2006 01:49:46 -0500 From: Joe Marcus Clarke <marcus@marcuscom.com> To: Doug Barton <dougb@FreeBSD.org> Cc: freebsd-gnome@FreeBSD.org Subject: Re: firefox 2 crashing with any printing-related activity Message-ID: <1166165386.15580.2.camel@shumai.marcuscom.com> In-Reply-To: <458239B5.8020103@FreeBSD.org> References: <4581C70F.10709@FreeBSD.org> <1166154994.40786.26.camel@shumai.marcuscom.com> <458239B5.8020103@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--=-ZOCth+w5K/5xzX6Dh9iD Content-Type: multipart/mixed; boundary="=-qjCNrf4lr0AQ5vTSowPi" --=-qjCNrf4lr0AQ5vTSowPi Content-Type: text/plain Content-Transfer-Encoding: quoted-printable On Thu, 2006-12-14 at 21:59 -0800, Doug Barton wrote: > Joe Marcus Clarke wrote: > > On Thu, 2006-12-14 at 13:50 -0800, Doug Barton wrote: >=20 > >> Howdy, > >> > >> For some time now, I have been having a problem with firefox 2. If I > >> do anything printing related (including just print preview) firefox > >> Seg faults. It will send the print job to the printer first, then > >> crash when the dialog closes. Here is the information you requested on > >> http://www.freebsd.org/gnome/docs/bugging.html > >> > >> FreeBSD lap 7.0-CURRENT-1213 FreeBSD 7.0-CURRENT-1213 #0: Wed Dec 13 > >> 21:36:15 PST 2006 i386 > >> > >> This is an Intel Core 2 Duo system running i386 SMP. I'm running with > >> X11BASE=3D/usr/local/xorg in order to help with the fixing of the port= s > >> prior to the move. > >> > >> pkg_info, env output, and gdb log are here: > >> > >> http://people.freebsd.org/~dougb/env.out > >> http://people.freebsd.org/~dougb/pkg_info.out > >> http://people.freebsd.org/~dougb/gdb.log > >=20 > > I can't reproduce. This stack trace is also missing four pretty > > critical frames. >=20 > Are you talking about 11-14? Are those frames, or are those stack values? >=20 > >> Ports are up to date as of this morning, base is up to date as of last > >> night. > >> > >> I tried this with CUPS off, and building firefox completely without > >> CUPS on the system at all, and the result is the same. I've also tried > >> building without using ccache. > >=20 > > This doesn't matter. Have you tried _running_ Firefox without cups-bas= e > > on the system?=20 >=20 > Well, yeah. Sorry if I wasn't clear. I've built firefox with and > without CUPS installed. On the last run I even moved the > libprintbackend-cups.* files out of the way to make sure they weren't > available, and verified that the only print option was > Postscript/default when I ran it. I also tried downloading the > package, and got the same result. >=20 > > CUPS will (once again) be dloaded by Firefox if present. > > I tried both with and without cups-base installed, and I still can't ge= t > > Firefox to crash. =20 >=20 > D'oh. Questions, are you using an SMP box, and are you using -current? > Because I can make a UP kernel, and/or try the test on RELENG_6 on the > same box if that'll help. >=20 > > I say, "once again" because we have seen a similar CUPS-related crash i= n > > the past, >=20 > I remember that, but it looks like this one is not CUPS' fault. >=20 > > and it looks like my bandaid fix was broken over the years. > > If it is the same problem that we saw before, removing cups-base during > > runtime should prevent the crash. If so, we can re-patch Firefox not t= o > > use libcups, and instead CUPS users can rely on the CUPS lpd commands t= o > > do their printing. It won't be as pretty, but it will at least avoid a > > crash (unless someone can finally track down what the conflict is > > between CUPS and Firefox). >=20 > Well, ugly + works > pretty + crashes. :) >=20 > Thanks to Jeremy's suggestion I created a new firefox with the debug > OPTION (duh, should have thought of that). I uploaded the following: >=20 > With CUPS > http://people.freebsd.org/~dougb/debug.log > http://people.freebsd.org/~dougb/gdb2.log >=20 > Without CUPS > http://people.freebsd.org/~dougb/debug-no-cups.log > http://people.freebsd.org/~dougb/gdb2-no-cups.log I believe I have found the problem. Firefox was double-freeing a font face. The attached patch should fix this. Joe --=20 PGP Key : http://www.marcuscom.com/pgp.asc --=-qjCNrf4lr0AQ5vTSowPi Content-Disposition: attachment; filename=patch-gfx_src_ps_nsFontMetricsPS.cpp Content-Transfer-Encoding: base64 Content-Type: text/x-patch; name=patch-gfx_src_ps_nsFontMetricsPS.cpp; charset=ISO-8859-1 LS0tIGdmeC9zcmMvcHMvbnNGb250TWV0cmljc1BTLmNwcC5vcmlnCUZyaSBEZWMgMTUgMDE6NDI6 MjUgMjAwNg0KKysrIGdmeC9zcmMvcHMvbnNGb250TWV0cmljc1BTLmNwcAlGcmkgRGVjIDE1IDAx OjQxOjIwIDIwMDYNCkBAIC0xMDM3LDggKzEwMzcsMTAgQEAgbnNGb250UFNYZnQ6OkluaXQobnNY ZnRFbnRyeSogYUVudHJ5LA0KIA0KIG5zRm9udFBTWGZ0Ojp+bnNGb250UFNYZnQoKQ0KIHsNCi0g IGlmIChtRW50cnktPm1GYWNlKSANCisgIGlmIChtRW50cnktPm1GYWNlKSB7DQogICAgIEZUX0Rv bmVfRmFjZShtRW50cnktPm1GYWNlKTsNCisgICAgbUVudHJ5LT5tRmFjZSA9IG5zbnVsbDsNCisg IH0NCiANCiAgIGlmIChGVF9Eb25lX0ZyZWVUeXBlKG1GcmVlVHlwZUxpYnJhcnkpKQ0KICAgICBy ZXR1cm47DQpAQCAtMjMyNyw4ICsyMzI5LDEwIEBAIG5zWGZ0VHlwZTFHZW5lcmF0b3I6OkluaXQo bnNYZnRFbnRyeSogYUUNCiANCiBuc1hmdFR5cGUxR2VuZXJhdG9yOjp+bnNYZnRUeXBlMUdlbmVy YXRvcigpDQogew0KLSAgaWYgKG1FbnRyeS0+bUZhY2UpIA0KKyAgaWYgKG1FbnRyeS0+bUZhY2Up IHsNCiAgICAgRlRfRG9uZV9GYWNlKG1FbnRyeS0+bUZhY2UpOw0KKyAgICBtRW50cnktPm1GYWNl ID0gbnNudWxsOw0KKyAgfQ0KIA0KICAgaWYgKEZUX0RvbmVfRnJlZVR5cGUobUZyZWVUeXBlTGli cmFyeSkpDQogICAgIHJldHVybjsNCg== --=-qjCNrf4lr0AQ5vTSowPi-- --=-ZOCth+w5K/5xzX6Dh9iD Content-Type: application/pgp-signature; name=signature.asc Content-Description: This is a digitally signed message part -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (FreeBSD) iD8DBQBFgkWKb2iPiv4Uz4cRAq0sAKCM4hF3duQ4STQyLn3fPt5tR2M1qgCdHSJt 3077X+ALRM6azgmyH8P8Emc= =vAxN -----END PGP SIGNATURE----- --=-ZOCth+w5K/5xzX6Dh9iD--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1166165386.15580.2.camel>