On Sunday 23 December 2007, Christian Perrier wrote: > For work in progress, that would allow translators to focus on the most > important strings and, for instance, to target completing sublevels 1 > and 2, before moving to the current "level 2" (tasksel, iso-codes...) > before coming back on "sublevel 3" later. Another option that was discussed and that would make inclusion of incomplete languages less of a problem, is to make localechooser display a warning if a translation is incomplete. This would allow the user to either select a different language if (s)he speaks one, or break off the installation if (s)he's not confident of being able to correctly interpret any untranslated messages at critical points in the translation. I've just committed an initial version of a script [1] that can generate the file to be used by localechooser to determine the correct warning (if any). To ensure that the status in the file really corresponds to the actual status of the translation in the release, the file retrieves the PO files for the version of all packages *in testing* (using the release tags in SVN). These are split into sublevel files and the translation statistics for each level are calculated. Here is a sample of the output of the script to the console: I: Gathering the data... I: processing anna (1.27) I: processing apt-setup (1:0.32) I: processing arch/alpha/aboot-installer (0.0.28) I: processing arch/arm/glantank (1.5) I: processing arch/i386/grub-installer (1.27) I: processing arch/i386/lilo-installer (1.25) [...] I: Calculating statistics... I: am is limited translated at 2: 432/1608 I: ar is mostly translated at 2: 1560/1608 I: be is mostly translated at 2: 1597/1608 I: bg is fully translated [...] I: ru is fully translated I: sk is mostly translated at 2: 1597/1608 I: sl is mostly translated at 2: 1554/1608 I: sq is mostly translated at 2: 1595/1608 I: sv is mostly translated at 2: 1606/1608 I: ta is mostly translated at 3: 227/234 [...] In the last lines the first number is the sublevel (sublevels 1 and 2 are combined as 2) and after the period are translated/total stings in the sublevel. Note that only a few udebs currently actually have the sublevel markers in the version in testing, so a lot of strings that should be in sublevels are included in level 1 instead. This should be resolved after Beta1. The output file currently looks as follows (this is an initial example and can be tuned when we implement the localechooser part): am: 2 L ar: 2 M be: 2 M bg: 5 F [...] ru: 5 F sk: 2 M sl: 2 M sq: 2 M sv: 2 M ta: 3 M [...] Cheers, FJP [1]http://svn.debian.org/wsvn/d-i/trunk/scripts/l10n/calc-release-status?op=file&rev=0&sc=0
Attachment:
signature.asc
Description: This is a digitally signed message part.