Apache OpenOffice (AOO) Bugzilla – Issue 15428
thencode: th_en_US hardcoded
Last modified: 2013-02-24 20:34:27 UTC
Kevin, we can not use thencode directly for Czech thesaurus database creation, because it has en_US hardcoded in it: pavel@pavel:~/Tmp/ooo_11beta2_src/lingucomponent/source/thesaurus/parser> grep th_en_US.dat * thencode.cxx: dpath = aoUrl + OUString( RTL_CONSTASCII_USTRINGPARAM("/th_en_US.dat" )); thencode.cxx: dpath = aoUrl + OUString( RTL_CONSTASCII_USTRINGPARAM("th_en_US.dat")); We must workaround it by mv $(BIN)/th_en_US.dat $(BIN)/th_cs_CZ.dat mv $(BIN)/th_en_US.idx $(BIN)/th_cs_CZ.idx It is OK for now, but I think it should be fixed by adding additional command-line argument to thencode (en_US or cs_CZ for us). What do you think?
Hi, Yes your solution is best. But I thought you were using a much simpler awk script and I was hoping you would "officially" contribute it and then thencode can go away for good. Also Giuseppe is completely rewriting thencode for the it_IT project because he had trouble with awk and "phrases with spaces in them". So I have high hopes that it_IT will contribute a much improved thencode for all of us to use. Until then I can add a parameter for which file name to use. Thanks, Kevin
Hi, Fixed in latest version of cws_srx644_ooo11beta2 New syntax is: thencode input_directory output_directory output_file_name So inside the newest dictionaries/en_US the makefile uses the following: ./thencode $(MISC) $(BIN) th_en_US which creates th_en_US.idx and the_en_US.dat So I am resolving this as fixed. Please verify in the next OOo 1.1 and close. Thanks, Kevin
I see another possible problem: thencode has another hardcoded values: wordlist.txt and trimthes.txt I added cs_CZ directory to dictionaries and after this, there already are those files thus $(MISC)$/wordlist.txt and $(MISC)$/trimthes.txt targets will not be rebuild and English thesauri effectively contains Czech words :-( I solved it by removing those files just after creating th_cs_CZ* files.
Hi Pavel, Good point! But removing at the end is not workable under parallel builds is it (try building en_US and cz_CZ in parallel!) Since I don't want to pass in yet more names perhaps we should change all of this yet again to look something like: thencode input_directory output_directory root_name So I would pass in something like ./thencode $(MISC) $(BIN) th_en_US and then thencode would read in th_en_US_words.txt th_en_US_thes.txt and produce th_en_US.idx and th_en_US.dat You would then pass in ./thencode $(MISC) $(BIN) th_cz_CZ and your input files would be th_cz_CZ_words.txt th_cz_CZ_thes.txt and your output files would be th_cz_CZ.idx th_cz_CZ.dat Would that be better? Kevin
This is one approach. BTW - I work on unix systems, thus names with 13 or more characters are OK, but is it OK in general? What about creating directory en_US or cs_CZ in $(MISC) and working with files wordlist.txt and trimthes.txt there?
Hi, I would rather go with unigue names like th_en_US_words.txt and the_en_US_thes.txt. I have already mande the liungucomponent and dictioanary changes to implement this in my own tree and tested it and it works fine (and will not interfere with parallel builds). Since dictionary.lst created no problems and it is 10.3 I would assume that 14.3 would be fine as well. But just to make sure, I have asked on the dev@porting and if 14.3 is fine, I will commit those changes. If that comes back as okay, I will commit these changes so I can close this once again! Thanks, Kevin
Kevin, I think we can go with long names.
I see it in 11RC :-) I have modified cs_CZ thesaurus parsing and it seems to work :-) I'm not able to test it in running RC, because it does not run - it only offers to repair or uninstall.
Found fixed in 11rc. Thanks Kevin.
-