On Tue, 20 Jan 2009 23:10:13 +0100 Jakub Wilk <uba...@users.sf.net> wrote:
> * Neil Williams <codeh...@debian.org>, 2009-01-20, 20:59: > >> > Perhaps we could have a tool converting .mo files from one endianess to > >> > the other and ship the two versions in epiphany-browser-data, > >> > >> Well, either msgfmt should be able to produce both or a special tool is > >> required. The later would be easier as this tool may fix the location at > >> the same time. > > > >msgfmt can do this already with --endian but it needs the original PO > >file which is in the source package. > msgunfmt can uncompile .mo files. Nicolas François has done some initial tests on a big endian machine. msgunfmt is a lot slower so isn't a runtime solution and doing a build time conversion means shipping two versions which merely duplicates all .mo files for an estimated 20 millisecond improvement per 10,000 strings. That leaves using msgunfmt at runtime which would make gettext a dependency of thousands of postinst scripts. Hmmm, not sure that's a good idea. All just to save ~20 milliseconds?? http://lists.debian.org/debian-devel/2009/01/msg00492.html Thanks for putting some figures to the discussion, Nicholas. I think the gettext wrapper is doing quite a good job, based on those figures. If anything, the figures could be used to support patching gettext to always output in big endian but that puts the "burden" onto a much larger number of users. True, those users generally have faster machines but I'm not sure everyone would like it. Can we drop this as an interesting diversion but a waste of time and effort to actually fix and let .mo files be architecture-independent again? -- Neil Williams ============= http://www.data-freedom.org/ http://www.nosoftwarepatents.com/ http://www.linux.codehelp.co.uk/
pgpMkzDMZutAd.pgp
Description: PGP signature
