Last modified: 2013-06-23 19:24:06 UTC
Looks like trailing whitespace does round trip. * {{ {{ucfirst:{{lc:UNSIGNED}}}} |user=Root|date=00:43, 21 May 2013 (UTC)}} -> * {{{{ucfirst:{{lc:UNSIGNED}}}} |user=Root|date=00:43, 21 May 2013 (UTC)}}
Are you sure this isn't a Parsoid bug? I've seen lots of template normalization issues to do with whitespace in Parsoid.
Confirmed that this is an issue in Parsoid.
If this template was modified, then normalizing the name should be fine. If the template was not modified then this is likely a VE DOM round-tripping bug.
Not a selser issue, this is Parsoid core issue; I'm OK with you WONTFIXing, but this is nothing to do with VE.
It is not a core issue. editMode is true on parsoid.wmflabs.org -- so all serialization in forms go through regular serialization. See below from commandline output: [subbu@earth tests] echo '* {{ {{ucfirst:{{lc:UNSIGNED}}}} |user=Root|date=00:43, 21 May 2013 (UTC)}}' | node parse --wt2wt * {{{{ucfirst:{{lc:UNSIGNED}}}} |user=Root|date=00:43, 21 May 2013 (UTC)}} [subbu@earth tests] echo '* {{ {{ucfirst:{{lc:UNSIGNED}}}} |user=Root|date=00:43, 21 May 2013 (UTC)}}' | node parse --wt2wt --editMode false * {{ {{ucfirst:{{lc:UNSIGNED}}}} |user=Root|date=00:43, 21 May 2013 (UTC)}} So, this is going through non-selser serialization which means that segment of the DOM is being marked dirty.
That just brings up an unrelated issue. Do we want editMode to be true / false on parsoid.wmflabs.org? I offhand dont remember the reason why we defaulted it to true there. Locally, my localsettings.js sets it to false.
I think that it should always be true, except in rt testing. That installation mainly serves ourselves, and seeing a fully serialized template including the minor diffs is useful in that context. Moving to VE again as the issue is most likely selser being disabled for the transclusion due to VE DOM modifications. Please test DOM round-tripping for this transclusion in VE.
Timo, can you clarify if this transclusion was modified by the user?
AFAICT this is an INVALID bug; marking as such. Timo, feel free to re-open if you can give context.