Alistair Crooks <agc%pkgsrc.org@localhost> writes: > And the counter argument goes: > > + make it difficult for people to commit, and no-one will commit I agree with a lot of your points in the abstract. But it's a question of degree. I didn't say anything (that I remember) in June and July, when breakage was at reasonable levels: http://releng.netbsd.org/b5reports/i386/ But in August, it has not been rare for breakage to last more than a few hours - by eyeball it looks like the tree was unbuildable (on i386) about 25% of the time. > + building is only the tip of the iceberg. What's next? complaints that > software doesn't work, fsvo "working" Well, there have been complaints about significant increases in the number of ATF failing test cases, which I think is a valid complaint. > + it is asking too much for people to be able to get things right 100% of > the time on all NetBSD platforms, what with unsigned chars vs signed chars, > 32 vs 64bit, big endian and little endian, etc. > > Some believe that -current should build at all times. I believe that that > view is not possible with the diverse set of areas in which people are > developing. I didn't mean to demand 100%. But I do think we should strive for better than 75%, with the result that many commits are made during periods of unbuildability (and therefore untestability for bisecting regresssions).
Attachment:
pgplE4b65CYCA.pgp
Description: PGP signature