https://bugs.freedesktop.org/show_bug.cgi?id=44135

--- Comment #59 from jwoithe <[email protected]> ---
In reply to Comment 56:
> Would Resolved->NotOurBug work for this ? [ of course, quite possibly it is 
> at least partially our bug of so not entirely satisfying ], if so, any chance 
> of closing it ? At least the details are archived here for future reference.

"NotOurBug" is probably not telling the whole story, although I'm not sure what
else to suggest since the resolution descriptions available are by design
fairly generic.  It is clear that a change in LO's PDF generation code broke
important applications like CUPS through LO versions 3.5, 3.6 and 4.0 (and
possibly earlier).  LO 4.1 changed something which made at least some versions
of CUPS render what was expected (see comment 53), but older versions of xpdf
(3.02 and earlier) - and perhaps other renderers too - still could not get the
render right.  Whether the 4.1 change was just a happy accident also remains to
be seen.  A noted in comment 51, CUPS 1.5.4 could not render output from LO 4.0
correctly while (as per comment 53) the same CUPS version on the same machine
could deal with the output from LO 4.1.  At the same time however, xpdf 3.02
could not render either output correctly.  Whatever change was made between LO
4.0 and LO 4.1, it would appear to have been subtle enough to make CUPS happy
while still creating issues for things like xpdf 3.02.  Furthermore, while we
don't know exactly what change made CUPS happy it is unclear whether the
altered behaviour is due to a targetted code change or is merely a side effect
of some other modification and therefore might break again later.

Personally I am a little uncomfortable with marking this as "NotOurBug" for
these and similar reasons to those outlined in comment 57.  While crxssi could
potentially pursue this though their RHEL support contract as suggested in
comment 58, I am sure there are many other users subject to locked down systems
from sources other than RHEL who would not have this option.  What makes the
situation somewhat tricky is that all affected PDF renderers used to work; it
was a change in LO which broke them.  At the same time, if LO's revised PDF
output is technically correct then perhaps the behaviour does point to bugs (or
missing feature implementations) in the affected renderers.  Checking the
changelog for xpdf 3.03 for example (http://www.foolabs.com/xpdf/CHANGES) it is
clear that a number of changes were made to transparency handling between 3.02
and 3.03 which explain why 3.03 may now have a better chance of getting it
right.

Having said all that, I have probably personally taken this as far as I have
time and knowledge to do.  I am not familiar with the LO codebase, nor have I
in-depth knowledge of the PDF format and its various implementations.  For all
machines which I manage (personally and at work) upgrading software is an
option and for us this seems sufficient to address the problem at least for
now.  As a result of all these things, spending more time on this is not
currently feasible for me.  If there is no one else to pick up the
investigation of the outstanding issues, then perhaps concluding it as
"NotOurBug" makes sense if we were willing to blame the problem solely on
deficiencies in other software.  However, looking back at the history this is
not overly satisfying as mentioned by Michael in comment 56.

-- 
You are receiving this mail because:
You are the assignee for the bug.
_______________________________________________
Libreoffice-bugs mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs

Reply via email to