I wrote: > We will need to change the CD build scripts around a bit now, so that > daily CDs are built (especially for non-i386 arches) that contain the > debs from testing, but the udebs from unstable (and appropriately > developmental versions of the initrds as well). This is so that we can > test that the udebs in unstable actually work before propigating them to > testing. Of course, we'll also want to keep the existing setup, so we > can build final CD images from testing, using the debs and udebs and > initrds in testing. I suppose we could call the former CDs daily, and > the latter CDs release candidates. > > Along similar lines, if you need to test a udeb in unstable, and you > netboot or use a boot floppy, d-i will currently use the udeb from > sarge. To force it to use the udebs from unstable, we will need to add a > boot time parameter. So due to an upload of a new upstream version of busybox-cvs-udeb to unstable, both of the above methods of testing are more or less impossible. You would be testing with the wrong version of busybox. At this point I don't know whether to -- a. Upload a new version of busybox that reverts it back to exactly what it was before. b. Do some horrendous hack to the CD scripts to use only specific packages (not busybox) from unstable, and the rest from testing. (This would not help netboot/floppy testing.) c. Not worry about it and let the porters deal with the resulting possible breakage and confusion themselves. d. Propigate the new busybox to testing for !i386, and let the porters deal with *that*, with possible arbitrary delays as the result. e. Release without any ports, since bleeding edge busybox is somehow more important. Needless to say, this sucks, and in case you didn't hear me the first time, UNSTABLE IS FROZEN TO UPLOADS OF UDEBS, EXCEPT FOR THOSE NECESSARY TO GET PORTS DONE. Sigh. -- see shy jo
Attachment:
signature.asc
Description: Digital signature