Andreas Gustafsson <gson%gson.org@localhost> writes:
> Bernd Ernesti wrote:
>
>> Anyway. I assume this error is related to a checkout which only got the
>> sbin part and not the sys one.
>
> Your assumption is correct, and the next build did succeed. Sorry about
> the false alarm. If only there was a way to mirror the repository that
> would guarantee a consistent snapshot...
For a second I thought I had messed up, even though I did a full build,
install and atf run before committing, and then updated and another
build...
While I agree on the theoretical problems about CVS and atomicity, most
10-minute intervals have no commits. So an algorithm:
redo:
start_time=gettimeofday()
do update
latest_time=`find latest time in checkout`
if (latest_time + 30s >= start_time)
goto red
that should avoid getting part of what is logically an atomic commit.
An alternative is to ignore this problem and only address it if the
false alarm rate is over 10%. I realize we are perhaps 1 for 1, and
that may be partly because people are trying harder to avoid triggering
the automatic reports.
Attachment:
pgpNSQ8qBlzu1.pgp
Description: PGP signature