Archive for May, 2005

holidays

Tuesday, May 24th, 2005

Still on honeymoon in South Africa, dumped out the safari pics. e.g. lion, buffalo, wild dogs, leopard, rhino, and elephant. Out in Camp Bay at the outskirts of Cape Town at the moment after travelling from Port Elizabeth over the last two weeks. Couple of more nights here and then back home. We drove over the Outeniqua Pass from Oudtshoorn a few days ago, on the approach it seems suspicously cloudy, when we rounded the bend we noticed that the mountain was on fire! News later that day was that the pass was closed shortly after we fought our way through the blaze

gone

Wednesday, May 4th, 2005

until May 28th

OOo stuff

Wednesday, May 4th, 2005

java/gcj
Behold! The ooo2 java wizards working under gcj . Now see what you’ve been missing. xsltfilters such as the docbook export and import filters also spring to life.

Due to gcc#19870# the wizards don’t just work out of the box with gcj. Crudely seding private and protected into public as these docbook and wizards patches do will workaround the gcj problem.

To get the wizards registered this patch is also required.

This gcj request asks for the addition of java.awt.Frame.createBufferStrategy which is all that is missing from gcj to build the java canvas stuff. (Though the canvas module contains a pile of spurious imports of sun.awt which are unnecessary and can be removed, not that there’s much point right now, if a createBufferStrategy becomes available then removing the sun.awt from the canvas/java .javas is all that’s outstanding)

Other build Stuff
The upstream libxmlsec still doesn’t have two of the symbols which the internal patched OOo libxmlsec contains and which the rest of OOo requires. A version of libxmlsec which provides these would allow a system-libxmlsec. *Hint Hint*

Frustratingly gcj and db-4 aren’t playing well together, ironically making me revert OOo’s internal db back to db3. Alternatively dropping the optimization for building the internal db4 from -O2 to -O1 seems to allow it to continue. More digging required.

Would be nice if someone familiar with both sablotron and libxslt would take a look at the one place in OOo where sablotron is used and see if a relatively trivial conversion to libxslt would be possible.

methinks the OOo environmental setting scripts should export SAL_ENABLE_FILE_LOCKING=1 setting this seems to have cleared a collision in cppumakers in parallel processes spawned by build.pl (not the multiple cppumakers spawned by parallel dmake btw). Even if that’s just dumb luck that it works for me it might at least work out some of the reported problems with file locking if it’s enabled for developers during their build :-)

rtf regression
Knocked out a fix for a regression in the rtf import which has apparently being floating around for a few years. Really hate that poxy rtf format. Rule of thumb for the OOo rtf filter is that whoever reads the { is responsible for reading the }