Control: tag 666670 + patch 2012-05-31 12:37 GMT+02:00 Jonas Smedegaard <[email protected]>: > severity 666670 important > thanks > > On 12-04-10 at 05:35pm, Agustin Martin wrote: >> I do not think this deserves serious priority, but would however leave >> an open bug report with lower priority asking for better efficiency if >> possible. > > Agreed. > > Thanks for commenting, Augustin - and for reporting, Lucas.
Hi, Jonas, Although a simple rebuild should fix #764209, I have been looking at how difficult would be to use autobuild in dsdo in relation to #764209. When I started looking at it I was hit by old #666670, dsdo: FTBFS: build hangs for at least 15 minutes so I looked at it and seems I have something interesting. As pointed out by Lucas, the problem is in > sort -u ../ispell/words-da.complete \ > | bash ../no-compound_marking ../words-da.no-compound \ > | LC_ALL=POSIX sort -u \ > > words-da.complete I have looked at ../no-compound_marking script and it seems the problem. It is a bash script with really poor efficiency, included in 1006_improve_myspell_rules.patch. I have been playing with a perl implementation for the same purpose and seems I have a working first cut of it. It is at least one order of magnitude faster than bash implementation and seems to really fix this bottleneck. While more testing is needed, it is creating exactly the same hunspell.dic file as old slow procedure, so it looks OK. I am attaching a patch with my current changes against sid dsdo. Since a new build will also fix #764209 I am cc'ing it. Also, it does not sort the input, sorting the output should suffice. I noticed that you already imported a new version 1.6.30 to the git repo and was playing a bit with it. To have my preferred look explicit merging of upstream into master seems missing, but that is a matter of personal taste. Also, seems that dpkg-source has become pedantic about old quilt patches, so old patches do not get properly applied (at least when using dpkg-source inside pdebuild) and failure happens. All those patches need refreshing, After that, things seem to work again. Hope this helps. If you are busy I can prepare an NMU with the changes you find convenient. Regards, -- Agustin
0001-Use-a-perl-script-to-handle-compound-flags.-Much-fas.patch
Description: application/download
