Inkscape crashes when I search the Open Clip Art Library

Reproduced with Inkscape 0.48.2 and 0.48.3 on OS X 10.7.2 (64bit)
Date: Fri Apr 27 01:55:37 CEST 2012

This must be due to a change in the RSS feed from openclipart: the search worked without crashing Inkscape 0.48.x just a a week ago (2012-04-20 according to irc log). Tonight, inkscape crashes when searching with 'pony' as well as with my standard search term when testing import from openclipart for reported issues ("apple").

@JazzyNico - any chance you could test with the new dialog in current trunk?

Crash report for Inkscape 0.48.3.1 OS X 10.7.2 (64bit)

> This must be due to a change in the RSS feed from openclipart: (…)

OpenCLipart seems to produce invalid feeds:
 Bug #988844 “Feeds are no longer valid”
 <https://bugs.launchpad.net/openclipart/+bug/988844>

Inkscape of course should not crash even when the feed is invalid - this report is closely related or a duplicate of the earlier ones, like e.g.
 Bug #889421 “Segfault in openclipart dialog when openclipart.org is down ”
 <https://bugs.launchpad.net/inkscape/+bug/889421>

Tested with Inkscape 0.48.2 and 0.48.3.1 on Ubuntu 11.04. No crash. Just get the following message in an error dialog (as expected):
"Failed to receive the Open Clip Art Library RSS feed. Verify if the server name is correct in Configuration->Import/Export (e.g.: openclipart.org)"

Also tested with the trunk. No error and no crash. The result list just stays blank when the throbber stops.

Reproduced on Debian Wheezy, Inkscape 0.48.1. Same crash report as the original reporter.
Additionnal gdb message (before bt):

-----
(gdb) run
Starting program: /usr/bin/inkscape
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/i386-linux-gnu/libthread_db.so.1".
[New Thread 0xb1025b70 (LWP 1966)]
http://openclipart.org/media/feed/rss/test:2: parser error : Extra content at the end of the document
<link rel="alternate" type="application/rss+xml" title="Top 10 Cliparts" href="/
^

Program received signal SIGSEGV, Segmentation fault.
0x08620b0b in Inkscape::UI::Dialog::FileImportFromOCALDialog::print_xml_element_names(_xmlNode*) ()
-----

No crash and no message with the trunk.

JazzyNico: It appears we're not trying to even get the feeds from the right place.

As opposed to:
http://openclipart.org/media/feed/rss/

We should be using:
http://openclipart.org/api/search/?query=

An example would be:
http://openclipart.org/api/search/?query=water

I did change the location locally in my copy of ocaldialogs.cpp locally and unfortunately it causes segfaults (however, when not modified it just fails quietly and doesn't crash).

For openclipart is because we move to new Aiki framework and it seams that there is bug in new version, but cann be fixed in this current version of Aiki, with one tweak.

I fixed on ocal side and I tested on Inkscape ver 0.48.2 on Ubuntu and it work fine. You can test it and mark for Inkscape as fixed.

PS: next time when you have this kind of bug please add Openclipart as another project that affect this bug.

> You can test it and mark for Inkscape as fixed.

Upstream fix in Openclipart confirmed on OS X 10.7.3 with Inkscape 0.48.2 and 0.48.3.1 ('File > Import from Open Clip Art Library' works fine again). Thx!

Proposing to keep the report open ('Confirmed') for project 'Inkscape' for now:
- As mentioned in comment #4, Inkscape ought to handle upstream issues with the feeds without crash (probably related to bug #889421).
- The report(s) need further testing & confirmation whether both issues are handled without causing a crash in current trunk (new ocal import dialog).

> PS: next time when you have this kind of bug please
> add Openclipart as another project that affect this bug.

Sorry to have missed that - will be done next time similar issues with ocal import in Inkscape are reported.

~ScislaC - Not sure about the link. Apparently, the generic API and the RSS link are almost identical. The generic API is still RSS, but adds a <content:encoded> element that encloses the description in CDATA.
And I see nowhere in the doc (http://openclipart.org/docs/api) which one should be preferred.

~Jakub - Could you please give you opinion on the API we should use in Inkscape?

JazzyNico: I got that from an email discussion with Jakub. When he asked about the url we were using, he gave the API link as an example. When I said we'd fix the url moving forward, there was no indication that we shouldn't or that there was no need. We can certainly wait to hear an official opinion though.

I rewrite the rss API recently because there where bugs, the xml was
invalid (there ware unescape & in text) I don't know if link that use
inkscape is free of this bugs and simple Inkscape it handle them.

I don't remember if RSS API is using Sphinx search (which is use for
normal search) if it doesn't then it will, so API search will be the
same as normal search you will get the same clipart. and sphinx is
faster and more accurate.

If they are almost the same then maybe I will need to remove one of
them, If I remove one link I will put redirect. Does Inkscape will
handle HTTP redirect?

Maybe Inkscape don't need to use new API for now. I don't want to break
old version of Inkscape so if Inkscape will not handle redirect I can
put the same code as for API search if it's the same purpose and return
the same stuff then it can be identical.

On Sat, 12 May 2012 18:20:52 -0000
JazzyNico <email address hidden> wrote:

> ~ScislaC - Not sure about the link. Apparently, the generic API and
> the RSS link are almost identical. The generic API is still RSS, but
> adds a <content:encoded> element that encloses the description in
> CDATA. And I see nowhere in the doc (http://openclipart.org/docs/api)
> which one should be preferred.
>
> ~Jakub - Could you please give you opinion on the API we should use in
> Inkscape?
>

--
Jakub Jankiewicz
www: http://jcubic.pl
twitter: http://twitter.com/jcubic
blog: http://jcubic.wordpress.com

Yes, keep compatibility and same feed URL are very important, no matter what.

Jon

On Sun, May 13, 2012 at 3:25 AM, Jakub Jankiewicz
<email address hidden> wrote:
> I rewrite the rss API recently because there where bugs, the xml was
> invalid (there ware unescape & in text) I don't know if link that use
> inkscape is free of this bugs and simple Inkscape it handle them.
>
> I don't remember if RSS API is using Sphinx search (which is use for
> normal search) if it doesn't then it will, so API search will be the
> same as normal search you will get the same clipart. and sphinx is
> faster and more accurate.
>
> If they are almost the same then maybe I will need to remove one of
> them, If I remove one link I will put redirect. Does Inkscape will
> handle HTTP redirect?
>
> Maybe Inkscape don't need to use new API for now. I don't want to break
> old version of Inkscape so if Inkscape will not handle redirect I can
> put the same code as for API search if it's the same purpose and return
> the same stuff then it can be identical.
>
>
> On Sat, 12 May 2012 18:20:52 -0000
> JazzyNico <email address hidden> wrote:
>
>> ~ScislaC - Not sure about the link. Apparently, the generic API and
>> the RSS link are almost identical. The generic API is still RSS, but
>> adds a <content:encoded> element that encloses the description in
>> CDATA. And I see nowhere in the doc (http://openclipart.org/docs/api)
>> which one should be preferred.
>>
>> ~Jakub - Could you please give you opinion on the API we should use in
>> Inkscape?
>>
>
> --
> Jakub Jankiewicz
> www: http://jcubic.pl
> twitter: http://twitter.com/jcubic
> blog: http://jcubic.wordpress.com
>
> --
> You received this bug notification because you are a member of
> openclipart.devel, which is subscribed to openclipart.
> https://bugs.launchpad.net/bugs/989257
>
> Title:
>  Inkscape crashes when I search the Open Clip Art Library
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/inkscape/+bug/989257/+subscriptions

--
Jon Phillips 王✳爻气 http://fabricatorz.com ✳ skype: kidproto ✳ irc: rejon
+1.415.830.3884 (global) ✳ +86-187-1003-9974 (beijing)

I've linked a similar/identical report in Debian. Feel free to unlink if you think this is unrelated.

> I've linked a similar/identical report in Debian.

Indeed Inkscape 0.48.3.1 is crashing once again when import from Open ClipArt (same build, no local changes).

@Jakub - have there been changes again in the feed? Would you prefer a new separate report? Unfortunately, I can't tell when importing from ocal in inkscape started to fail again.

Attaching strace from Debian tracker, if this is of any use.

I don't know what happen, I update Aiki yesterday, but this url look fine http://openclipart.org/media/feed/rss/cat and validator show it's valid rss

Content-Type is application/xml; should it be different? Ah and I enable gzip on the server (but Strace show xml tags). Does those changes can be the cause of this?

Inkscape was killed by this line:

<content:encoded><![CDATA[ Cartoon cat. ]]></content:encoded>

I remove it

Cool, Josh, can you make sure Inkscape list and people find out that
openclipart search should be working again.

Thanks!

On Wed, Jun 6, 2012 at 5:34 AM, Jakub Jankiewicz
<email address hidden> wrote:
> Inkscape was killed by this line:
>
> <content:encoded><![CDATA[ Cartoon cat. ]]></content:encoded>
>
> I remove it
>
> ** Changed in: openclipart
>       Status: Confirmed => Fix Committed
>
> --
> You received this bug notification because you are a member of
> openclipart.devel, which is subscribed to openclipart.
> https://bugs.launchpad.net/bugs/989257
>
> Title:
>  Inkscape crashes when I search the Open Clip Art Library
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/inkscape/+bug/989257/+subscriptions

--
Jon Phillips 王✳爻气 http://fabricatorz.com ✳ skype: kidproto ✳ irc: rejon
+1.415.830.3884 (global) ✳ +86-187-1003-9974 (beijing)

Jakub Jankiewicz wrote
> Inkscape was killed by this line:
> <content:encoded><![CDATA[ Cartoon cat. ]]></content:encoded>
> I remove it

Thx for investigating and figuring out the actual trigger: upstream fix confirmed with Inkscape 0.48.2 and 0.48.3.1 on OS X 10.7.4 (searching and importing from Open Clipart works again without Inkscape crashing).