fix pbs static ldif paths when GIP_CE_USE_CACHE is true

[pigay]

Workaround for #135

[jouvin]

I am reluctant to accept this fix. See comment in #135.

[pigay]

I don't understand your hesitation.

It remains that the config file refers to a file which is not configured by quattor. I couldn't find any piece of a template where the symlinks are create. This must have disappeared from the repo at some point.

The people involved into upgrading to the new lcg-info-dynamic-scheduler will probably have to refactor the gip/ce/pan template, so it doesn't harm to accept this fix.

Anyway. Feel free to reject this PR. I just intended to contribute a bit.

[jouvin]

@pigay read the GGUS ticket, it explains why this file should be outside of the standard location. The real fix would probably be the other way around: ensure that the file is created in the right location. Because of the bug in the GGUS ticket, it was temporarily created in the only location where it could be found by the buggy version of lcg-info-dynamic-scheduler. Now that the pb is fixed and the config file taken into account, I think we should create it at the right place rather than amend the config file (your PR). I hope to find the time to work on it this afternoon...

[jouvin]

Except if you want to do it!

[pigay]

@jouvin I had missed the GGUS part indeed and didn't understand your comment on #135.

I'd be very happy to try to help if I could. But I'm even unable to guess how these static files are generated.

[jouvin]

I started working on it... In fact, as I was remembering, I left FIXME lines explaining what had to be done so should not be too difficult... More later!

[jouvin]

As explained, this workaround is not the right fix... Replaced by #141.

[jouvin]

As said in #141, let's merge this one into 15.4 as a temporary workaround until the filxed dynsched-generic is really released (end of May), giving time for more testing and clearing the last issues reported by @pigay in #141 discussion.