Debian Bug report logs - #1100748
segfault when using get_text() on arm64

Reply or subscribe to this bug.

Display info messages

Message #5 received at [email protected] (full text, mbox, reply):

From: Johannes Schauer Marin Rodrigues <[email protected]>

To: Debian Bug Tracking System <[email protected]>

Subject: segfault when using get_text() on arm64

Date: Tue, 18 Mar 2025 07:05:50 +0100

Package: python3-pymupdf
Version: 1.25.0+ds1-2
Severity: serious

Hi,

the autopkgtest of pymupdf currently segfaults on arm64, ppc64el and
riscv64. In an attempt to get to the bottom of this I found a minimal
reproducer:

python3 -c "import pymupdf,sys; doc=pymupdf,sys; doc=pymupdf.open(sys.argv[1]);doc[0].get_text()" /usr/share/doc/fig2dev/manual.pdf

Which pdf document is opened seems to be irrelevant. It crashed for any that I
tried. This makes this bug RC, I think.

I can reproduce the problem in a minimal Debian unstable chroot:

mmdebstrap --include=python3-pymupdf,python3-pymupdf-dbgsym,libmupdf25.1-dbgsym,gdb,fig2dev --customize-hook='env SHELL=/bin/sh chroot "$1" gdb -batch -ex "set logging enabled on" -ex "set pagination off" -ex run -ex "bt full" -ex quit --args python3 -c "import pymupdf,sys; doc=pymupdf,sys; doc=pymupdf.open(sys.argv[1]);doc[0].get_text()" /usr/share/doc/fig2dev/manual.pdf' unstable /dev/null http://deb.debian.org/debian "deb http://deb.debian.org/debian-debug unstable-debug main"
[...]
Program received signal SIGSEGV, Segmentation fault.
0x0000fffff2d13ce8 in mupdf::FzStextPageIterator::FzStextPageIterator (this=0xfffff74b5e6c, item=...) at platform/c++/implementation/classes.cpp:22939
warning: 22939	platform/c++/implementation/classes.cpp: No such file or directory
#0  0x0000fffff2d13ce8 in mupdf::FzStextPageIterator::FzStextPageIterator (this=0xfffff74b5e6c, item=...) at platform/c++/implementation/classes.cpp:22939
No locals.
#1  0x0000fffff2d13d3c in mupdf::FzStextPage::begin (this=this@entry=0x11dd690) at platform/c++/implementation/classes.cpp:22868
        ret = {m_item = {m_internal = 0x6665727800000000, static s_num_instances = 1}}
#2  0x0000fffff74ad020 in JM_print_stext_page_as_text (res=..., page=...) at ./src/build/extra.i.cpp:5570
        block = {m_internal = 0xfffff7fe6420}
        __for_range = @0x11dd690: {m_internal = 0x1196ea8}
        __for_begin = {m_item = {m_internal = 0xfffff7fe6430}}
        __for_end = {m_item = {m_internal = 0xffffffffe7c0}}
        rect = {x0 = <optimized out>, y0 = 0, x1 = <optimized out>, y1 = 792}
#3  0x0000fffff74ad320 in _wrap_JM_print_stext_page_as_text (self=<optimized out>, args=<optimized out>) at ./src/build/extra.i.cpp:10767
        fail = <optimized out>
        resultobj = 0x0
        arg1 = 0x11dd560
        arg2 = <optimized out>
        argp1 = 0x11dd560
        res1 = <optimized out>
        argp2 = 0x11dd690
        res2 = <optimized out>
        swig_obj = {0xfffff196af90, 0xfffff196b380}
#4  0x00000000004e924c in ?? ()
No symbol table info available.
#5  0x00000000004a3dc0 in _PyObject_MakeTpCall ()
No symbol table info available.
#6  0x00000000004c1fd4 in _PyEval_EvalFrameDefault ()
No symbol table info available.
#7  0x00000000005d2040 in PyEval_EvalCode ()
No symbol table info available.
#8  0x00000000005fd91c in ?? ()
No symbol table info available.
#9  0x00000000005f7e70 in ?? ()
No symbol table info available.
#10 0x00000000005e6094 in ?? ()
No symbol table info available.
#11 0x00000000005e5e1c in ?? ()
No symbol table info available.
#12 0x00000000006229f8 in Py_RunMain ()
No symbol table info available.
#13 0x00000000005bc188 in Py_BytesMain ()
No symbol table info available.
#14 0x0000fffff7ce229c in ?? () from /lib/aarch64-linux-gnu/libc.so.6
No symbol table info available.
#15 0x0000fffff7ce237c in __libc_start_main () from /lib/aarch64-linux-gnu/libc.so.6
No symbol table info available.
#16 0x00000000005badb0 in _start ()
No symbol table info available.
A debugging session is active.

	Inferior 1 [process 4068] will be killed.


Any idea what could be going on?

Thanks!

cheers, josch

Message #12 received at [email protected] (full text, mbox, reply):

From: Johannes Schauer Marin Rodrigues <[email protected]>

To: [email protected], [email protected], [email protected], [email protected]

Subject: Re: Bug#1100748: Acknowledgement (segfault when using get_text() on arm64)

Date: Mon, 21 Apr 2025 19:06:13 +0200

Control: reassign -1 mupdf 1.25.1+ds1-5
Control: tag -1 + patch

Hello mupdf and pymupdf maintainers,

my package plakativ is not in testing because of an RC bug in pymupdf. 
It turns out that to fix #1100748, src:mupdf has to be changed like 
this:

--- a/debian/rules
+++ b/debian/rules
@@ -36,7 +36,7 @@ else
     CFLAGS += -O2
 endif

-BUILD_FLAGS = build=debug shared=yes OS=Linux XCFLAGS="$(CFLAGS)" 
XLDFLAGS="$(LDFLAGS)"
+BUILD_FLAGS = build=debug shared=yes OS=Linux XCFLAGS="$(CFLAGS)" 
XLDFLAGS="$(LDFLAGS)" XCXXFLAGS="-DNDEBUG"
 BUILD_FLAGS += LINUX_OR_OPENBSD=yes
 # Force using system library
 BUILD_FLAGS += USE_SYSTEM_MUJS=yes

For details, please see the comments of Julian Smith: 
https://github.com/pymupdf/PyMuPDF/issues/4390

mupdf maintainers: could you take care of uploading a new version of 
src:mupdf with above change? If you don't have time, I can offer to NMU 
src:mupdf for you.

As far as my testing goes, pymupdf seems to work fine even without being 
rebuilt. I wonder if we should schedule a binNMU after the src:mupdf 
fix. Bastian, what do you think?

I brought up this topic in #debian-release today and Sebastian Ramacher 
said that since pymupdf is in Bookworm, this would be eligible for an 
exception. Bastian, do you want to file the pre-approval bug or would 
you like me to take care of it?

Thanks!

cheers, josch

P.S.: currently on vacation and hence sending without PGP signature, 
sorry

Message #25 received at [email protected] (full text, mbox, reply):

From: Daniel Echeverri <[email protected]>

To: Johannes Schauer Marin Rodrigues <[email protected]>

Cc: [email protected], [email protected], [email protected]

Subject: Re: Bug#1100748: Acknowledgement (segfault when using get_text() on arm64)

Date: Mon, 21 Apr 2025 13:52:37 -0500

[Message part 1 (text/plain, inline)]

Hello Josch!

El lun, 21 abr 2025 a la(s) 12:15 p.m., Johannes Schauer Marin Rodrigues (
[email protected]) escribió:

> Control: reassign -1 mupdf 1.25.1+ds1-5
> Control: tag -1 + patch
>
> Hello mupdf and pymupdf maintainers,
>
> my package plakativ is not in testing because of an RC bug in pymupdf.
> It turns out that to fix #1100748, src:mupdf has to be changed like
> this:
>
> --- a/debian/rules
> +++ b/debian/rules
> @@ -36,7 +36,7 @@ else
>       CFLAGS += -O2
>   endif
>
> -BUILD_FLAGS = build=debug shared=yes OS=Linux XCFLAGS="$(CFLAGS)"
> XLDFLAGS="$(LDFLAGS)"
> +BUILD_FLAGS = build=debug shared=yes OS=Linux XCFLAGS="$(CFLAGS)"
> XLDFLAGS="$(LDFLAGS)" XCXXFLAGS="-DNDEBUG"
>   BUILD_FLAGS += LINUX_OR_OPENBSD=yes
>   # Force using system library
>   BUILD_FLAGS += USE_SYSTEM_MUJS=yes
>
> For details, please see the comments of Julian Smith:
> https://github.com/pymupdf/PyMuPDF/issues/4390
>
> mupdf maintainers: could you take care of uploading a new version of
> src:mupdf with above change? If you don't have time, I can offer to NMU
> src:mupdf for you.
>
> As far as my testing goes, pymupdf seems to work fine even without being
> rebuilt. I wonder if we should schedule a binNMU after the src:mupdf
> fix. Bastian, what do you think?
>
> I brought up this topic in #debian-release today and Sebastian Ramacher
> said that since pymupdf is in Bookworm, this would be eligible for an
> exception. Bastian, do you want to file the pre-approval bug or would
> you like me to take care of it?
>
> Thanks!
>
> cheers, josch
>
> P.S.: currently on vacation and hence sending without PGP signature,
> sorry
>

Thanks for your report and the patch!  I will work on it tonight.

Regards!

-- 
Daniel Echeverri
Debian Developer
Linux user: #477840
GPG Fingerprint:
D0D0 85B1 69C3 BFD9 4048 58FA 21FC 2950 4B52 30DB

[Message part 2 (text/html, inline)]

Message #30 received at [email protected] (full text, mbox, reply):

From: Debian FTP Masters <[email protected]>

To: [email protected]

Subject: Bug#1100748: fixed in mupdf 1.25.1+ds1-6

Date: Tue, 22 Apr 2025 03:05:02 +0000

[Message part 1 (text/plain, inline)]

Source: mupdf
Source-Version: 1.25.1+ds1-6
Done: Daniel Echeverri <[email protected]>

We believe that the bug you reported is fixed in the latest version of
mupdf, which is due to be installed in the Debian FTP archive.

A summary of the changes between this version and the previous one is
attached.

Thank you for reporting the bug, which will now be closed.  If you
have further comments please address them to [email protected],
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Daniel Echeverri <[email protected]> (supplier of updated mupdf package)

(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing [email protected])


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

Format: 1.8
Date: Mon, 21 Apr 2025 17:12:48 -0500
Source: mupdf
Architecture: source
Version: 1.25.1+ds1-6
Distribution: unstable
Urgency: medium
Maintainer: Kan-Ru Chen (陳侃如) <[email protected]>
Changed-By: Daniel Echeverri <[email protected]>
Closes: 1100748
Changes:
 mupdf (1.25.1+ds1-6) unstable; urgency=medium
 .
   * Fix segfault when using get_text() on arm64 (Closes: #1100748)
     + Thanks to Johannes Schauer for the patch
   * Update symbols file
Checksums-Sha1:
 e0c71fa42bd43e84ce86d4f3bf60e94534a25065 2661 mupdf_1.25.1+ds1-6.dsc
 f616d2d6d010149ea52929afad18ac8a7cb58d1c 94216 mupdf_1.25.1+ds1-6.debian.tar.xz
 e80e7d3ff61b4c338a4e2ce1be6ab247565f41b9 16163 mupdf_1.25.1+ds1-6_amd64.buildinfo
Checksums-Sha256:
 afc3e3dacd9e04c427e7535bbc06e971d886a5d316373c82e83d26e8e342fa02 2661 mupdf_1.25.1+ds1-6.dsc
 1c799488ad82e1655be122004f0664cca5ce1034dad073a77c90c0a5f54d365b 94216 mupdf_1.25.1+ds1-6.debian.tar.xz
 166b9b781c7e65ccd44f0571af2397a0a8dcbc1008401b40c6304614bd92e810 16163 mupdf_1.25.1+ds1-6_amd64.buildinfo
Files:
 3b55e0f97442f52870c9b56da0566b2e 2661 text optional mupdf_1.25.1+ds1-6.dsc
 bdf5ce2ba81e61f1d7d2d15f72a0d2b9 94216 text optional mupdf_1.25.1+ds1-6.debian.tar.xz
 c9a3de99f8d080acddee3a8ff222a3cc 16163 text optional mupdf_1.25.1+ds1-6_amd64.buildinfo

-----BEGIN PGP SIGNATURE-----

iQJHBAEBCgAxFiEE0NCFsWnDv9lASFj6IfwpUEtSMNsFAmgHBD8THGVwc2lsb25A
ZGViaWFuLm9yZwAKCRAh/ClQS1Iw2zznD/9W7DhtV8uxjkAwUe6Y4wPCrZlmAJiY
DxjpJtE2wXWsTE9c55fgVOqTZ0JrlPtLAW/4DPfe8BEsykLW8IUz7JK1fDf8Tybm
lURdSqBnxXh/878y86TG8XhdcZzBOOig3xw5ol2hBGsDawn8k1GPEfo/XJMCBmxM
HR2TT0UgDjAgoIcgtDdpEtP75Pa8Z0a69c3S3bCWQhwTz4q4WRu7EU5OLO+AK7Xd
0JLoNSx5bHJ5XPPWGSMDjCLkGunvA+9fImXlgVxjMYMyFSSSdM8jBvxHdCO7s+WC
sy9+qf305eo/0ZWSb4vYjfwtXlyNhkDBtJC0/YE009Djugd0hDX+yLbhMTXbTV5j
u9TUx2TsfsGYrqEhF5FxQlKhW606xlRtva3iPQKJkBJW4ibJ3f0Wz3rU6MWj8QTQ
QAQQE2arEfI+VXUBIch9THOfnDTMaR0rTTcPslaoC3GGadyAe16aZbBfyKFRmzMi
TqFLWrenaZMmudqekUoGebRmV5yu14mG49GqGqTTVdMJLJ/D5aV6KWpL1Dj17OXr
PmAPZo4WjkFV/FKOZwqCOCkeMk9WA3o+XhJoE7x6UBM0VXi4gGq86AHtCvkakgOH
200IGqPRQG64dTXjWiAemzW4uLW1DANedhYKksJoAMGfonImsgavn/TztTGKPymT
u0YZUM3uECIclg==
=rrbE
-----END PGP SIGNATURE-----

[Message part 2 (application/pgp-signature, inline)]

Send a report that this bug log contains spam.