On Sunday 15 February 2009, Otavio Salvador wrote: > Frans Pop <elendil@planet.nl> writes: > > On Sunday 15 February 2009, Otavio Salvador wrote: > >> The script itself is not yet (it still lacks debian-installer > >> building) but it already does a good job for rest of packages. > > > > Have you ever looked at scripts/buildscript? > > Yes but mine uses cowbuilder to do all the work also tries to catch the > architecture compatibility by itself. > > As I told, it is not yet ready but is already a good start and once I > have time again I plan to finish it. > > What cons you see about the use of cowbuilder? Why should I tell you about what cons I see? I've never even used cowbuilder myself. Why don't you instead tell me what's so wrong with 'buildscript' that it needs to be re-implemented? You have a somewhat strange way of asking for comments/help: - you don't say what your goals are - you don't say what problems you're encountering that you need help with (if any) One reason I did not want to discuss this privately is that I've never been convinced we should do this at all, for a number of reasons. So I'm really not all that motivated to look at your script at all, especially if there are no specific questions. Let me give the reasons why I don't think it is a good idea: - ftp-masters are already not happy with source compliance for the daily builds [1]; this would be *a lot* worse for builds from SVN - builds from SVN are a lot more likely to be broken than daily builds and it would need to be distinguished from all the builds we already have when users report bugs; so: extra confusion and who the hell is going to do the extra work that will result from that? - if we have pending changes that are ready, someone should just take the responsibility for the upload (either a committer or someone else; IMO taking responsibility for the upload does however also include at least basic testing and staying alert for regressions) - for getting translation updates, instead of building from SVN someone should just do more frequent uploads [2] - it is extremely optimistic to expect translators to suddenly start testing installs just because their latest string changes are included The *only* valid reason I can see to do a build from SVN is when a new language has been activated and the relevant translator wants to see how it looks. But in that case such a build can much better be done on an ad-hoc basis. Cheers, FJP [1] This came up during the meeting at FOSDEM; I don't fully agree with them, but still. [2] Of course I DO NOT mean an upload every time the French translation team adds a cedille below a "c" (i.e: c -> ç), but if there are a significant number of translation updates pending for a package, that's a valid reason to do an upload.
Attachment:
signature.asc
Description: This is a digitally signed message part.