Date: Tue, 4 Aug 2020 11:06:43 +0200 From: Niclas Zeising <zeising+freebsd@daemonic.se> To: Ashish SHUKLA <ashish@FreeBSD.org>, Kevin Oberman <rkoberman@gmail.com> Cc: "freebsd-x11@freebsd.org" <x11@freebsd.org>, emacs@freebsd.org Subject: Re: Recent libX11 update broke emacs Message-ID: <6646fd04-f7c7-7e9e-99f6-f6ceeddaf888@daemonic.se> In-Reply-To: <6684a1e6-862c-51da-ce17-9fafdf23b2f4@FreeBSD.org> References: <CAN6yY1sOsGd5YdjAMdO01opmDvV8LRcqPQUg3gze5yu0XOKShQ@mail.gmail.com> <3fd8de9f-5746-1216-6047-af576d851b1a@daemonic.se> <15394c3c-27ba-5548-45ee-ea1e98b56a27@FreeBSD.org> <6684a1e6-862c-51da-ce17-9fafdf23b2f4@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2020-08-04 06:48, Ashish SHUKLA wrote: > On 8/4/20 10:04 AM, Ashish SHUKLA wrote: >> On 8/4/20 3:38 AM, Niclas Zeising wrote: >>> On 2020-08-03 19:07, Kevin Oberman wrote: >>>> This is a weird problem and I'm uncertain if it is in libW11, > x11-server, >>>> or emacs. As a result, I'm crossposting to both lists. >>>> >>>> Yesterday, after upgrading libX11 and x11-server, I could no longer = run >>>> Emacs to an X display. It crashed quickly with a segmentation error >>>> in the >>>> pthread library. Oddly, I could run it to my local xserver from a re= mote >>>> system. >>> >>> Where does the segfault occur before entering pthreads? >>> >>>> Rolling back the x11 server did not help nor did rolling back libX11= or >>>> even both of them. I had to roll back libX11 first and build the ser= ver >>>> with the old libX11. Rebuilding emacs with the new ports made no >>>> difference. >>> >>> Which version of the server did you build? The patch to xorg-server = is >>> trivial, just switching out malloc for calloc have the memory 0 >>> initialized. The patch to libX11 is rather more involved. >>> It is possible that someone accidentally changed the ABI of libX11, I >>> guess. What happens if you use the version with just the security fi= xes >>> (the version I committed first, and that was merged to the Q3 branch.= ) >>> >>>> >>>> Since the old libX11 contains a very serious vulnerability, I reall= y >>>> hate >>>> to leave it this way, but I also have a hard time getting anything d= one >>>> without emacs. Nothing other than emacs seems to have a problem with > this >>> >>> There is one commit after libX11 1.6.10, that takes care of a regress= ion. >>> Can you apply >>> > https://gitlab.freedesktop.org/xorg/lib/libx11/-/commit/93fce3f4e79cbc7= 37d6468a4f68ba3de1b83953b >>> on top of the libX11 1.6.10 update and see if that fixes anything? >>> >>> Regards >> >> Able to get Emacs to segfault: >> >> =CE=BB lldb -e /usr/local/bin/emacs >> (lldb) target create "/usr/local/bin/emacs" >> Current executable set to '/usr/local/bin/emacs' (x86_64). >> (lldb) run -Q >> Process 36196 launching >> Process 36196 launched: '/usr/local/bin/emacs' (x86_64) >> Process 36196 stopped >> * thread #1, name =3D 'emacs-28.0.50', stop reason =3D signal SIGSEGV: >> invalid address (fault address: 0x0) >> frame #0: 0x0000000000427ca0 >> emacs`best_xim_style(xim=3D0x0000000000000000) at xfns.c:2629:21 >> (lldb) bt > [...] >=20 >> >> >> And workaround is launching Emacs like: env -u XMODIFIERS emacs >> >> I'll try the libX11 patch and give it a shot. >=20 > Emacs, with libX11 built with the patch, works fine now. It seems like > it's also reported in Debian[1], and they merged it. Maybe we can do th= e > same ? >=20 > References: > [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D966691 >=20 Thank you, I'll see about adding that patch to the port and package then. It looks like this is both a bug in libX11 and to some extent also emacs? Kevin, can you double check that the patch works for you. I find it=20 strange that you have to rebuild both xorg-server and libX11 for it to=20 work, are you forcing static linkage somewhere? Regards --=20 Niclas
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6646fd04-f7c7-7e9e-99f6-f6ceeddaf888>