Re: [BUG] can't compile semantic
Mats Lidell writes:
>>>>> Stephen wrote:
Stephen> OK, but I've found that sometimes distclean isn't enough, and
Stephen> somebody else recently reported that distclean wasn't enough.
Stephen> Then, if you once get a build that succeeds, it becomes
Stephen> impossible to replicate the bug.
So you are suggesting that the smoketest should try harder to start
from a clean state in order to find more problems?
I don't know what this implies for the smoke test. I would argue that
the smoke test probably should start from a clean slate to find
*fewer* problems. A problem that goes away by building in a fresh
checkout probably is not something that should hold up the smoketest,
since the workaround is easy but fixing the problem may not be.
Rather, we want the smoketest to find problems that prevent a build
even in a fresh checkout, and get them fixed as fast as possible.
Also, the smoketest is a peculiar beast that rebuilds on almost every
checkin. So it's likely to have a different state from everybody
else, but only you have access to that state. So if it catches this
kind of hard-to-reproduce bug, debugging it will be a burden on you,
personally. If you think that's a good use of your time, I don't
mind. ;-)
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-beta