_html_splitlines reopens wrong tag in certain cases

[Tim Hatch <trac@…>]

When run on the output of the PHP highlighter (php -sn), _html_splitlines chokes. This is due to having two different passes for opening and closing passes. It pretty much comes down to an incorrect interpretation of

<span class="x">
</span><span class="y">

The span open on the next line should be y, not x. This is solved by using a single regexp to walk the source, and keeping a more rigid stack of tags that are open. This is something that I've done before and am willing to code up a fix in a couple of days.

I'm attaching a patch to add a testcase in trac/mimeview/tests/api.py​ for this issue. This is actually a real-world issue for the following sample, when using the php-specific highlighter:

<?php
/*
comment
*/
?>

[Tim Hatch <trac@…>]

Added test for api.py

[Tim Hatch <trac@…>]

First attempt at fix, diff against trunk@3423

[Tim Hatch <trac@…>]

My first attempt​ passes the tests in tests/api.py along with a couple others I added (the more important one is in mimeview-html-splitlines-close-order.diff​ above) which were previously failing.

I am using this on my copy of Trac locally and have not noticed any regressions in the display of rendered files.

I'm open to comments — it's not as pretty as it could be (especially wrt reversing the open_tags list twice).

[Tim Hatch <trac@…>]

Proof of bug as exhibited on t.e.o

[Tim Hatch <trac@…>]

Bump.

I just noticed this bug affects t.e.o with the Enscript highlighter for Python. See trunk/trac/__init__.py​ — the whole license should be highlighted as a string, but reverts to p_identifier instead of p_tripledouble. The SilverCity highlighter doesn't trigger this bug on this specific file, but Enscript does. Here's a sanity test and one for the specific issue raised on t.e.o: attachment:mimeview-html-splitlines-bugtest.diff​

This is fixed by attachment:mimeview-html-splitlines-close-order-fix.diff​ above. Also, it's been long enough that 'devel' now refers to 0.11-dev, so it should probably retargeted at 0.10 since it shows up on t.e.o.

[Tim Hatch <trac@…>]

Turns out I got highlighter names backwards. Triggered with SilverCity, not Enscript.

[Tim Hatch <trac@…>]

As of r4399, this ticket should be no longer relevant. Is this 'invalid' or 'wontfix' at this point?

[sid]

Not completely intuitive yet, but this would probably be a worksforme based on TracTicketTriage guidelines.

[Christian Boos]

Well, taking all the versions into account (as discussed in #4298) we have for this ticket:

Affects Version: 0.10.3	Fixed in Version: — (wontfix)	(1)
Affects Version: 0.11	Fixed in Version: 0.11 (fixed)	(2)

1. because it's not worth the trouble, I think
2. fixed by being replaced by the new code (r4399)

As the reported version here is 0.10.3, I think wontfix is appropriate (well, unless someone really insists this fix goes into stable).