Build farm log

[Eduardo Lima-6]

Hi,

I'm starting a work of tracking Samba4 build errors (on PowerPC
architecture) using Build Farm and report them in bugzilla. I have some
questions and I would appreciate if someone can explain some things to me.
Suggestions are welcome too.

The latest PPC build done until this time is fd4061d (build Farm
link<http://build.samba.org/?function=View+Build;host=tridge;tree=samba_4_0_test;compiler=gcc>
).
Build status is: ok/ok/ok/ok

What does this status mean?

Error log field shows several errors. Some of these errors are in the make
field and others in the test field. Why is the status of make field showing
"passed"? All the tests have to pass to the build be considered green?

Do you have any recommendation about opening a bug report? It would be great
if someone else that is doing this kind of work could briefly explain how he
is doing it, thus we can follow the same good practices.

Thanks and regards,

--
Eduardo Lima

Hi Eduardo,

 > The latest PPC build done until this time is fd4061d (build Farm
 > link<http://build.samba.org/?function=View+Build;host=tridge;tree=samba_4_0_test;compiler=gcc>
 > ).
 > Build status is: ok/ok/ok/ok
 >
 > What does this status mean?

It means it is passing all the tests that it is expected to pass.

 > Error log field shows several errors. Some of these errors are in the make
 > field and others in the test field. Why is the status of make field showing
 > "passed"? All the tests have to pass to the build be considered green?

During the build the compiler produces quite a few warnings on
stderr. Most of these warnings don't matter, although the warnings do
make it harder to spot real errors. Some of the warnings may also be
important. Occasionally developers put some effort into reducing the
number of warnings (such as what Matthias has been doing with const
warnings), but this can be surprisingly difficult to get right in some
cases.

Some of the error output comes from runtime during 'make test'. This
has several causes. One cause is tests that we currently expect to
fail (those listed in source/selftest/knownfail). Those are not
considered errors for the overall result, but can produce quite a bit
of output. Otherwise are due to warnings in the code that are
non-fatal. For example, the ndr_pull_error warnings may be produced by
code which deliberately sends invalid RPC packets.

 > Do you have any recommendation about opening a bug report? It would be great
 > if someone else that is doing this kind of work could briefly explain how he
 > is doing it, thus we can follow the same good practices.

I don't think a big report is appropriate for these errors, unless you
have a specific error in mind which you don't think should be
happening.

Cheers, Tridge