Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
Quoting Pino Toscano (2021-01-03 01:43:04) > In data domenica 27 dicembre 2020 03:53:01 CET, Jonas Smedegaard ha scritto: > > Version: 9.53.3~dfsg-6 > > > > Quoting Pino Toscano (2020-12-22 10:08:12) > > > In data lunedì 21 dicembre 2020 18:23:12 CET, Simon McVittie ha scritto: > > > > > On my side, rebuilding libspectre1 solved this on my system. > > > > > > > > If a simple rebuild with no source changes fixes the symptoms of > > > > a bug, that might indicate an unintended ABI break in libgs9, or > > > > perhaps a bug in the old libgs9 headers (but fixed in the new > > > > headers) in code that gets inlined into libspectre at compile > > > > time. > > > > > > Both of them are issues in ghostscript anyway. > > > > This was fixed in Ghostscript since release 9.53.3~dfsg-6 - I just > > forgot to mention it in changelog (that will be corrected in next > > release). > > Oh nice, I did not notice it. I can confirm that using > - libgs9 9.53.3~dfsg-6 > - libspectre1 0.2.9-1 > - evince 3.38.0-3 > there are no problems opening PS files in evince. Thanks for the confirmation! Happy new (gregorian) year, - Jonas -- * Jonas Smedegaard - idealist & Internet-arkitekt * Tlf.: +45 40843136 Website: http://dr.jones.dk/ [x] quote me freely [ ] ask before reusing [ ] keep private signature.asc Description: signature
Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
In data domenica 27 dicembre 2020 03:53:01 CET, Jonas Smedegaard ha scritto: > Version: 9.53.3~dfsg-6 > > Quoting Pino Toscano (2020-12-22 10:08:12) > > In data lunedì 21 dicembre 2020 18:23:12 CET, Simon McVittie ha scritto: > > > > On my side, rebuilding libspectre1 solved this on my system. > > > > > > If a simple rebuild with no source changes fixes the symptoms of a > > > bug, that might indicate an unintended ABI break in libgs9, or > > > perhaps a bug in the old libgs9 headers (but fixed in the new > > > headers) in code that gets inlined into libspectre at compile time. > > > > Both of them are issues in ghostscript anyway. > > This was fixed in Ghostscript since release 9.53.3~dfsg-6 - I just > forgot to mention it in changelog (that will be corrected in next > release). Oh nice, I did not notice it. I can confirm that using - libgs9 9.53.3~dfsg-6 - libspectre1 0.2.9-1 - evince 3.38.0-3 there are no problems opening PS files in evince. Thanks! -- Pino Toscano signature.asc Description: This is a digitally signed message part.
Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
Version: 9.53.3~dfsg-6 Quoting Pino Toscano (2020-12-22 10:08:12) > In data lunedì 21 dicembre 2020 18:23:12 CET, Simon McVittie ha scritto: > > > On my side, rebuilding libspectre1 solved this on my system. > > > > If a simple rebuild with no source changes fixes the symptoms of a > > bug, that might indicate an unintended ABI break in libgs9, or > > perhaps a bug in the old libgs9 headers (but fixed in the new > > headers) in code that gets inlined into libspectre at compile time. > > Both of them are issues in ghostscript anyway. This was fixed in Ghostscript since release 9.53.3~dfsg-6 - I just forgot to mention it in changelog (that will be corrected in next release). Thanks to all involved in triaging this bug, - Jonas -- * Jonas Smedegaard - idealist & Internet-arkitekt * Tlf.: +45 40843136 Website: http://dr.jones.dk/ [x] quote me freely [ ] ask before reusing [ ] keep private signature.asc Description: signature
Processed (with 1 error): Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
Processing commands for cont...@bugs.debian.org: > reassign 975387 libgs9 ghostscript/9.53.0~dfsg-1 Bug #975387 [libspectre1] libspectre1: evince hangs with PS files, is there a need to rebuild it? Bug #977754 [libspectre1] evince does not display EPS or PS files anymore and shows "Loading..." forever Bug reassigned from package 'libspectre1' to 'libgs9'. Bug reassigned from package 'libspectre1' to 'libgs9'. No longer marked as found in versions libspectre/0.2.9-1. No longer marked as found in versions libspectre/0.2.9-1. Ignoring request to alter fixed versions of bug #975387 to the same values previously set Ignoring request to alter fixed versions of bug #977754 to the same values previously set Bug #975387 [libgs9] libspectre1: evince hangs with PS files, is there a need to rebuild it? Bug #977754 [libgs9] evince does not display EPS or PS files anymore and shows "Loading..." forever Marked as found in versions ghostscript/9.53.0~dfsg-1. Marked as found in versions ghostscript/9.53.0~dfsg-1. > retitle 975387 wrong size check for display_callback_v2_s struct Bug #975387 [libgs9] libspectre1: evince hangs with PS files, is there a need to rebuild it? Bug #977754 [libgs9] evince does not display EPS or PS files anymore and shows "Loading..." forever Changed Bug title to 'wrong size check for display_callback_v2_s struct' from 'libspectre1: evince hangs with PS files, is there a need to rebuild it?'. Changed Bug title to 'wrong size check for display_callback_v2_s struct' from 'evince does not display EPS or PS files anymore and shows "Loading..." forever'. > severity 975387 serious Bug #975387 [libgs9] wrong size check for display_callback_v2_s struct Bug #977754 [libgs9] wrong size check for display_callback_v2_s struct Severity set to 'serious' from 'important' Severity set to 'serious' from 'important' > forwarded 975387 https://bugs.ghostscript.com/show_bug.cgi?id=703301 Bug #975387 [libgs9] wrong size check for display_callback_v2_s struct Bug #977754 [libgs9] wrong size check for display_callback_v2_s struct Set Bug forwarded-to-address to 'https://bugs.ghostscript.com/show_bug.cgi?id=703301'. Set Bug forwarded-to-address to 'https://bugs.ghostscript.com/show_bug.cgi?id=703301'. > tag 975387 + patch Bug #975387 [libgs9] wrong size check for display_callback_v2_s struct Bug #977754 [libgs9] wrong size check for display_callback_v2_s struct Added tag(s) patch. Added tag(s) patch. > merge 975387 975574 Bug #975387 [libgs9] wrong size check for display_callback_v2_s struct Bug #977754 [libgs9] wrong size check for display_callback_v2_s struct Unable to merge bugs because: forwarded of #975574 is '' not 'https://bugs.ghostscript.com/show_bug.cgi?id=703301' severity of #975574 is 'normal' not 'serious' affects of #975574 is '' not 'evince' Failed to merge 975387: Did not alter merged bugs. > thanks Stopping processing here. Please contact me if you need assistance. -- 975387: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=975387 975574: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=975574 977754: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=977754 Debian Bug Tracking System Contact ow...@bugs.debian.org with problems
Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
reassign 975387 libgs9 ghostscript/9.53.0~dfsg-1
retitle 975387 wrong size check for display_callback_v2_s struct
severity 975387 serious
forwarded 975387 https://bugs.ghostscript.com/show_bug.cgi?id=703301
tag 975387 + patch
merge 975387 975574
thanks
In data lunedì 21 dicembre 2020 18:23:12 CET, Simon McVittie ha scritto:
> > On my side, rebuilding libspectre1 solved this on my system.
>
> If a simple rebuild with no source changes fixes the symptoms of a bug,
> that might indicate an unintended ABI break in libgs9, or perhaps a bug
> in the old libgs9 headers (but fixed in the new headers) in code that
> gets inlined into libspectre at compile time.
Both of them are issues in ghostscript anyway.
The rebuild, while "easy as it seems", does not consider an important
fact. From what I see, libgs got new APIs in 9.53.0, and libspectre
does not (optionally) use any of them. This means that a rebuild most
probably would not cause libspectre to require the newer ghostscript,
migrating instantly to testing. Considering that what we have here
smells like a behaviour change, this means making libspectre unusable
for the users of testing, and I do not find this acceptable.
I checked the changes between ghostscript 9.52.1 and 9.53.3, and I found
one thing: gs 9.53.0 reworks a bit the display device stuff, adding a
v3 device_callback struct, and keeping the support for what is now v2:
http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=eed3bad23510e59278bdaa5f7d0ab01fc1a1c21b;hp=04e937862eaa7e66bb9a87109874112cd354bf6f
display_callback is actually used in libspectre, see spectre-device.c.
Because it relies on DISPLAY_VERSION_MAJOR/DISPLAY_VERSION_MINOR, this
explains why it works after a rebuild, as it will use the
device_callback v3. This also shows that a rebuild is a no-no, as it
will get in the situation I described earlier (i.e. stop working with
ghostscript in testing).
There is actually code in ghostscript to support the old device_callback
(v2, as built in libspectre during the last built), as it can be see in
devices/gdevdsp.c, function display_check_structure:
static int display_check_structure(gx_device_display *ddev)
[...]
else if (ddev->callback->size == sizeof(struct display_callback_v2_s)) {
/* V2 structure with added display_separation callback */
if (ddev->callback->size != sizeof(display_callback))
return_error(gs_error_rangecheck);
[...]
Considering that:
- sizeof(struct display_callback_v2_s) != sizeof(display_callback);
the addition to new stuff to display_callback was the reason for the
new struct in the first place, so of course the two structs do not
have the same size
- the last libspectre build uses display_callback v2, so the check:
ddev->callback->size == sizeof(struct display_callback_v2_s)
is true
then the check "ddev->callback->size != sizeof(display_callback)" will
be always false! Indeed, tried a simple patch to drop this, and evince
shows again PS files without rebuilding libspectre.
I submitted this as ghostscript bug:
https://bugs.ghostscript.com/show_bug.cgi?id=703301
Because of this, I'm reassigning 977754/975387 to ghostscript, merging
also 975574 to them, and setting the proper metadata. I'm also attaching
a copy of the patch I submitted upstream.
Thanks,
--
Pino Toscanodiff --git a/devices/gdevdsp.c b/devices/gdevdsp.c
index 0a66a0278..52087f8d6 100644
--- a/devices/gdevdsp.c
+++ b/devices/gdevdsp.c
@@ -1430,9 +1430,6 @@ static int display_check_structure(gx_device_display *ddev)
}
else if (ddev->callback->size == sizeof(struct display_callback_v2_s)) {
/* V2 structure with added display_separation callback */
-if (ddev->callback->size != sizeof(display_callback))
-return_error(gs_error_rangecheck);
-
if (ddev->callback->version_major != DISPLAY_VERSION_MAJOR_V2)
return_error(gs_error_rangecheck);
signature.asc
Description: This is a digitally signed message part.
Processed (with 2 errors): Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
Processing control commands: > reassign 977754 libspectre 0.2.9-1 Bug #977754 [evince] evince does not display EPS or PS files anymore and shows "Loading..." forever Bug reassigned from package 'evince' to 'libspectre'. No longer marked as found in versions evince/3.38.0-3. Ignoring request to alter fixed versions of bug #977754 to the same values previously set Bug #977754 [libspectre] evince does not display EPS or PS files anymore and shows "Loading..." forever There is no source info for the package 'libspectre' at version '0.2.9-1' with architecture '' Unable to make a source version for version '0.2.9-1' Marked as found in versions 0.2.9-1. > merge 975387 977754 Bug #975387 [libspectre1] libspectre1: evince hangs with PS files, is there a need to rebuild it? Unable to merge bugs because: package of #977754 is 'libspectre' not 'libspectre1' Failed to merge 975387: Did not alter merged bugs. > affects 975387 + evince Failed to mark 975387 as affecting package(s): failed to get lock on /srv/bugs.debian.org/spool/lock/975387 -- Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. Unable to lock /srv/bugs.debian.org/spool/lock/975387 Resource temporarily unavailable at /usr/local/lib/site_perl/Debbugs/Common.pm line 692. at /usr/local/lib/site_perl/Debbugs/Common.pm line 650. -- 975387: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=975387 977754: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=977754 Debian Bug Tracking System Contact ow...@bugs.debian.org with problems
Re: Bug#977754: evince does not display EPS or PS files anymore and shows "Loading..." forever
Control: reassign 977754 libspectre 0.2.9-1 Control: merge 975387 977754 Control: affects 975387 + evince On Mon, 21 Dec 2020 at 14:58:55 +0100, Patrice Duroux wrote: > I suggest you to look at #975387 as I think it is related to this issue. Reassigning to libspectre and merging with the libspectre bug: this looks like something lower-level than evince itself. > On my side, rebuilding libspectre1 solved this on my system. If a simple rebuild with no source changes fixes the symptoms of a bug, that might indicate an unintended ABI break in libgs9, or perhaps a bug in the old libgs9 headers (but fixed in the new headers) in code that gets inlined into libspectre at compile time. smcv
