The FreeDict source package builds the TEI XML source files, representing the
dictionary,into compiled dictionaries in the dict format. It uses XSL
stylesheets for this. Some of the dictionaries have a considerable complexity
(> 150 MB of size) and the stylesheets are known to match in huge depth of the
XML syntax tree. This leads to a vast consumption of RAM and slow processing,
currently somewhere over 16G of RAM and building times of the whole package of
6-16 h, depending on the machine and the number of parallelism. The more
concurrency is used, the higher the RAM requirements will get.
This can be only solved by a systematic rewrite of the stylesheets for which
help would be appreciated upstream.
For the time being, this bug report serves as a reminder on the Debian side
that this issue still persists.

Thanks

Hi

Is there anything to instruct the build machines to allow the "excessive" RAM
usage of around 16G (maybe a bit more)? Is there any control field which would
allow this?
The package didn't FTBFS since a while now and I'll close this bug likely
soon.

Cheers
Sebastian

Hi,

Not that I'm aware of.
infrastructure, and this bug is listed as the cause. Maybe there is
another cause.

Example log (not a stable URL unfortunately):
https://reproduce.debian.net/all/api/v0/builds/296089/log

All logs can be found here:
https://reproduce.debian.net/all/stats/

Chris