Genshi since 0.5.1 produces duplicated items in py:match

[felix.schwarz@…]

Since the update to Genshi 0.5.1 the content from most of our py:match directives appears twice in the output. This happens on all platforms (Linux, Windows). I verified that this is still a problem in the current 0.5 stable branch (r908).

[felix.schwarz@…]

This is a test case for the described problem (can reproduce it with the current 0.6 trunk, too)

[felix.schwarz@…]

The change of behavior was introduced in r884, r885.

[felix.schwarz@…]

Actually, I found another problem in our code which was caused by the mentioned commits. However, for this second problem there is a workaround by using //h1 as match expression (see the second testcase)

[felix.schwarz@…]

Ok, I found a workaround for us now: The duplication test case does pass if the first py:match (which is essentially '<div py:match="div">') is modified to use '<div py:match="body//div">'.

So it's essentially the same workaround as for the "no match" test case.

[felix.schwarz@…]

Sometimes I had to resort to something like '*/div'. For now this is okay with us but it seems to be very fragile so we are eager to see a real fix for this.

[cmlenz]

I'm too tired for grokking the details of what's going on here, but here two simple options for fixing the first example you posted:

• The once="true" attribute can only be used when py:match is used as an element. In your example, it's simply ignored and passed on through into the output. If you make that match template a <py:match> example, the duplication goes away.
• Moving the first match template after the other two match templates also gets rid of the duplication.

[felix.schwarz@…]

Thanks for looking at this. Unfortunately, we can't change the order of the py:match directives as we're re-using some trac templates and all these py:matches originate from trac. Only the last one with the wrong 'py:match' is ours.

[didley@…]

I'm using 0.5.2dev and I got this in my admin panel

Could be it's the same reason.

didley

[didley@…]

Screenshot

[stromnov@…]

Is it fixed in 0.6dev branch?

[Felix Schwarz <felix.schwarz@…>]

No, both tests still fail for trunk (as of r1059).

[osimons]

Isn't this the same as #293?

[Felix Schwarz <felix.schwarz@…>]

I don't think so because I checked some weeks ago with the current trunk at that time but the bug was still there.

[Felix Schwarz <felix.schwarz@…>]

Just as a note for further investigation: This bug might be present in 0.5.1 already. We had a  customer report that indicated adding a site.html in trac (hence a fourth py:match) exposed the same behavior.

[Felix Schwarz <felix.schwarz@…>]

finally a patch for the dreaded issue

[cboos]

Great! I don't pretend to understand the fix, but it seems to work.

However import re is missing, for the test:

genshi/template/tests/directives.py

@@ -12 +12 @@
 # history and logs, available at http://genshi.edgewall.org/log/.
 
 import doctest
+import re
 import sys
 import unittest

Otherwise all tests pass on Windows, if that helps. I'll keep the change in my test environment for Trac to check if there are no adverse effects.

[cmlenz]

Patch applied in r1093. Thanks you so much!

[Felix Schwarz <felix.schwarz@…>]

Thanks for applying the patch :-)

You won't release another 0.5.x, right? Otherwise I would like to see the patch also for this branch - if the regression will make it in a stable release, we'll have a lot of support troubles.

[anonymous]

i tried to change the condition in setup.py file from 0.6 ---> 0.5 and it works fine