Hello all,
a short weekly update from my side. I was not that motivated this week.
1) Bazel experiment:
The build still crashes, but i see a slow improvement. It crashes now in
the user wizard.
I have migrated extensions, extras and added fonds, default_images,
ooo_custom_images and readlicense_so to complete the requirements at
startup.
currently i hardcode the language in the staging folder. The productive
build would delay this until the install. I have a 3 phase plan how to
improve on this including an migration from pootle to sdf Files.
I added following fix:
# Force MSVC iterator-debugging OFF for ALL modules so the STL
# container ABI is identical in debug and release builds. In
# release _HAS_ITERATOR_DEBUGGING already defaults to 0; in a /MDd
# _DEBUG build it defaults to 1, which adds a _Container_proxy to
# std::vector/hash_map and changes their layout. Several modules
# (comphelper, tools, svtools, sw, oox, …) already set =0 locally
# because they don't even compile at =1 (heterogeneous comparators);
# any module that DOESN'T set it then disagrees on container layout,
# so an object built in one DLL (e.g. comphelper::SequenceAsHashMap)
# and read inline in another (e.g. desktop/spl FirstStart) indexes a
# garbage bucket → "vector subscript out of range" assert / crash.
# Matches upstream solenv (iterator-debugging off even in debug).
# Symbols/PDBs/-Od are unaffected. Makes the per-module
# /D_HAS_ITERATOR_DEBUGGING=0 copies redundant.
"/D_HAS_ITERATOR_DEBUGGING=0",
@Damjan, Arrigo
is this something that makes sense? Is this a good solution or something
to revisit?
I am uncertain here.
2) extension Website
no progress, and on hold at the moment.
3) pootle
i checked what i need to do to set it up, and was not very motivated to
start.
i hope i can change that next week, and at least have a local install
and load a pootle file.
Am 23.05.2026 um 11:52 schrieb [email protected]:
Hi all,
I am writing every week a short summary, what is going on in my tasks.
Last week i wrote 2 mails to the list which were topic specific.
This week i dont manage to do this, it seems i loose fokus a bit. So
maybe it makes more sense to write a weekly checkin.
1) Bazel experiment
OpenOffice is still crashing. I figured that the main reason are the
ini files. Originaly we build them in scp2, with bash scripts and
generators.
I thought scp2 is building the installer, and did not want to have a
packaged object before i know that it also runs. That is why i wrote
that I changed the progress.
Turns out that our generators are not easy to predict by AI, and it
produced AI Slop ini files, Which kept me busy.
I migrated now scp2 and i hope that it is a step forward. The
migration was not that streight forward. I have now more AI wired
decisions. First we had a lot of bash, now i moved all bash to perl
and tried to reuse existing perl scripts from solenv.
Now i am curious if the direction is better. But once i have a working
version i need to review the complete build. There are multiple points
now that work, but are not ideal. The Progress slows now the closer I
get to the milestone.
2) extension Website
not much progress to report. I am currently moving the icon set. but
overall time for AOO was littlle not so murch progress.
3) Pootle
Drew pinged me on infra translation ticket, what the situation is for
the machine.
I claimed i put the migration backwards on priority due to other topics.
We need more disk space on oppenoffice-ec2-va, because Pootle and Open
grok need a lot of space. And the configuration is one disk.
In order to fullfil the transformation ticket, i thought about quickly
" Copy everything over. I am not sure if this is overall the right move.
moving to weblate is the move. But a migration over machines and
services has more risk that it needs more time.
I dont think i can do 3 topics in parallel. I will have to push back
the Extension for now, in favour of the translation migration.
Lets see. I try to work next week on the bazel experiment and on pootle.
All the best
Peter.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
