Mailinglist Archive: opensuse-buildservice (145 mails)

< Previous Next >
Re: [opensuse-buildservice] submitting a specific revision does not work

Quoting Olaf Hering <olaf@xxxxxxxxx>:

On Mon, Oct 21, Dominique Leuenberger a.k.a. Dimstar wrote:

Quoting Olaf Hering <olaf@xxxxxxxxx>:
>What does it check to verify if rev N is good? I see its been 6 hours
>between 242 and 243. To me it looks like something never attempts to
>built what was actually submitted. Are you saying OBS just waits for
>results in the origin, instead of building the submitted, different
>sources once more? If so, that smells like a error.

The verification is rather simple:
- Does rev N as submitted build

This is probably what should happen, but its not whats happening today.

It is what is checked: but it relies on OBS history data and does not go out and build stuff.. Just imagine the waste of time for LO => the script would be 'waiting' for 6 hours.. and so does the OBS worker building it.
This sounds like a duplication of tasks for the rare benefit of this special case where the maintainer submitted, changed and submitted; expecting the old SR to be valid.

rebuilding stuff on submit does not sound like the best plan neither.. The
main question should probably be: why was there a rev 243 checked in and why
would this not be important enough to be forwarded (well in fact it WAS
forwarded, but the request was declined (by me) for not following the
packaging guidelines).

IF this request would have been accepted, what would have been the point of
rev 242? Are we just having a meta-discussion here?

I dont think so. Its not so much about this specific package.

It is a corner case which happens once in a while.. investing additional workers dedicated to this corner case does not sound like the right thing to do. Better educate the prj maintainers about this.

BTW: the reason is always the same: prj maintainer changed the package after submitting, but before successful builds. In most of those cases, the prj maintainer ever would want the latest modification in as well (so, submitting HEAD as recommended would be the right thing)

If 'osc sr -r N' does not work, whats the point of having it? Perhaps
the checks for Factory are just incomplete. As I said, if submitted rev
does not match HEAD, then the checker should use and build "submitted
rev" instead of looking at something that was not submitted.

osc sr -r N DOES work (against all random projects). simply for Factory there is the additional check: did -r N ever successfully build.

The checker really ought to do what it's name sais 'check'. Not build.. building is a waste of power, if you get a full stack of 500 packages submitted. Throwing workers at it would mean other packages don't have workers available..

All in all: it's a corner case imho that is not worthy the effort of special build power.

Dominique
--
To unsubscribe, e-mail: opensuse-buildservice+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse-buildservice+owner@xxxxxxxxxxxx

< Previous Next >