| <<<Back 1 day (to 2015/09/06) | 20150907 |
_wiz_ | hello again. any mupdf people around? | 09:25.36 |
tor8 | _wiz_: yes. | 09:25.52 |
_wiz_ | hello, tor8! | 09:26.12 |
| the latest mupdf release doesn't build against the latest jbig2dec release. | 09:26.24 |
tor8 | which operating system? | 09:26.47 |
_wiz_ | it tries to include a non-existing memento.h | 09:26.49 |
| the fix is to add "-DJBIG_NO_MEMENTO" to mupdf's configure args. | 09:27.04 |
| I would presume "all of them", but I see it on NetBSD. | 09:27.13 |
tor8 | are you building from tarballs or a git clone? | 09:27.26 |
_wiz_ | mupdf-1.7a release tarball, jbig2dec-0.12 tarball | 09:27.41 |
tor8 | the release tarball comes prepackaged with a jbig2dec directory that it has been tested against | 09:28.20 |
_wiz_ | that's true, but when packaging software it's good to use other packages to avoid having multiple copies of the same (possible vulnerable) code on the system | 09:28.52 |
kens | Just sent a mail to Joann copied to tech, I'm feelign awful today so I'm taking a sick day. I will occasionally check the IRC logs and email so if you really, really need me try that. But be aware I'm not good for much today...... | 09:29.17 |
tor8 | yes, unfortunately you then run an untested configuration which may have other security issues of its own :) | 09:29.22 |
Robin_Watts | kens: I hope you feel better soon. | 09:29.34 |
_wiz_ | tor8: that might be, but I assume in general there'll be less of those :) | 09:29.45 |
kens | Robin_Watts : so do I, this started on Friday :-( | 09:29.57 |
kens | heads back to bed | 09:30.03 |
Robin_Watts | _wiz_: Normally I'd suggest whinging at the the people responsible for jbig2dec and/or memento :) | 09:30.32 |
_wiz_ | good point. I'll do that too. | 09:31.57 |
tor8 | _wiz_: I just did a build of mupdf-1.7a-source with the jbig2dec directory replaced with the jbig2dec 0.12 release | 09:32.05 |
| and that had no issues | 09:32.10 |
Robin_Watts | _wiz_: You miss the point. I wrote memento, and we are responsible for jbig2dec as well :) | 09:32.38 |
_wiz_ | robin_watts: heh | 09:32.48 |
| ok :) | 09:32.49 |
| tor8: jbig2dec does not install the memento.h header | 09:33.07 |
tor8 | so the jbig2dec-0.12 release is what's broken? nothing to do with mupdf? | 09:33.40 |
Robin_Watts | jbig2dec is primarily used for gs and mupdf. As such the memento.h header lives in those packages. | 09:33.58 |
| so the jbig2dec-0.12 release should set JBIG2_NO_MEMENTO by default, I think. | 09:34.22 |
tor8 | Robin_Watts: it looks like jbig2.h (the public header) includes memento.h | 09:34.29 |
| but the 'make install' target for jbig2dec never copies memento.h into /usr/include | 09:34.45 |
Robin_Watts | That sounds broken. | 09:34.56 |
| Possibly it should only be in the private header? | 09:35.10 |
| Unless there are Memento_label calls ? | 09:35.19 |
tor8 | so using jbig2dec 0.12 as a source package is never going to work | 09:35.28 |
| that is, if I have understood the problem correctly | 09:35.45 |
| yeah, memento should probably be in the private header | 09:36.08 |
chrisl | I thought that was tried and didn't work..... | 09:36.25 |
tor8 | or we should rename it jbig2_memento.h and install that | 09:36.41 |
_wiz_ | btw, the jbig2dec release distfile is broken in another way too: | 09:36.45 |
| it installs symlinks to the autotools files instead of copies | 09:36.54 |
| installs -> includes | 09:37.00 |
chrisl | _wiz_: that's fixed now | 09:37.05 |
_wiz_ | e.g. jbig2dec-0.12/compile -> /usr/share/automake-1.14/compile | 09:37.14 |
chrisl | Robin_Watts: http://git.ghostscript.com/?p=jbig2dec.git;a=commit;h=91744cfb2f and http://git.ghostscript.com/?p=jbig2dec.git;a=commit;h=5e5cff776abeb | 09:38.10 |
_wiz_ | chrisl: fixed how? you need to provide a new distfile to fix that. | 09:38.11 |
Robin_Watts | chrisl: Damn, I can't remember any details for that. | 09:39.03 |
chrisl | _wiz_: I fixed the options passed to automake | 09:39.10 |
| _wiz_: http://git.ghostscript.com/?p=jbig2dec.git;a=commitdiff;h=3837ea91 | 09:39.27 |
_wiz_ | chrisl: looks good, thanks. | 09:40.02 |
chrisl | _wiz_: BTW, do you have *any* clue as to why creating symlinks is the default? That seems totally insane to me | 09:40.34 |
_wiz_ | chrisl: no idea. I agree it's not a good default. | 09:41.13 |
| perhaps they assume everyone just runs 'autoreconf -fi' | 09:41.35 |
chrisl | Given these tools are supposed to make it easier to build everywhere, defaulting to a environment specific setup is just mental :-( | 09:42.24 |
_wiz_ | I feel your pain... | 09:42.41 |
chrisl | Personally, I wouldn't use automake..... | 09:43.02 |
_wiz_ | as a packager: please do use it. most alternatives are much worse | 09:43.30 |
| (cmake is not so bad) | 09:43.42 |
chrisl | I'd prefer just to use autoconf - I hate the way automake hides so much | 09:44.25 |
| Robin_Watts: I don't really understand the comment in the "revert" commit above | 09:45.34 |
_wiz_ | chrisl: what does it hide that you want to see? | 09:48.12 |
chrisl | Robin_Watts: in fact, given that jbig2dec ought to be using callbacks to allocate/free memory through the gs memory manager(s), I'm not clear why we'd need memento in jbig2dec at all | 09:48.21 |
| _wiz_: I can't recall off hand, I just remember finding it terribly frustrating the last time I had to mess with it | 09:48.57 |
| back in a while...... | 10:15.35 |
Robin_Watts | mvrhel_laptop: Morning. | 16:19.06 |
| I reckon you're looking for some sort of fz_measure_string type thing. | 16:19.22 |
| I was going to say that we don't have such a beast, but I figure we must as part of the html layout stuff. | 16:23.55 |
| but that would be a question for tor8. | 16:24.01 |
| Forward 1 day (to 2015/09/08)>>> | |