[opam-devel] dose/opam patches WAS [Re: ows reports]
louis.gesbert at ocamlpro.com
Thu Oct 2 09:59:41 BST 2014
Yes, I should have submitted them earlier. More details on why the need for them arose:
1. That one may not be strictly needed anymore: at some point we wanted to check the solver version by first trying with advanced criteria, then falling back to more basic ones in case of error. The issue was that after the solver call failed, raising an exception, the timer would still be set, which would make the next call fail with this assert.
2. Hm, really should have done this earlier, I'll need to look it up but I _did_ get to a case where I wanted more info on the conflicts and this check prevented me from getting the conflict information. I'll find you the precise case.
3. OPAM already includes functions for running external commands, with logging, signal handling, etc. It's nicer from our side to be able to call the solver command consistently with the same functions and error handling. The different way to handle signals already proved a problem a couple times. Also, we were trying to achieve more compatibility between different solvers, and adding a layer at that point may get handy ; this could also be used to send a query to the solver farm, without any external command.
On Wednesday 01 October 2014 12:01:40 Pietro Abate wrote:
> On 01/10/14 10:13, Roberto Di Cosmo wrote:
> > On Tue, Sep 30, 2014 at 09:56:00PM +0100, Anil Madhavapeddy wrote:
> > > David Sheets points me to the version of Dose packaged up in OPAM as the
> > > right version to use here (it has various patches applied to it for
> > > compatibility with opamLib, I believe).
> > Thanks for reminding me of these patches: none is opam specific: two seem to fix
> > real issues in dose, and one adds an extra function.
> I've reviewed the three patches Roberto sent to me this morning.
> - "Remove broken assert"
> I merged this one. Do you have a quick example ? I can imagine that
> this could happen, but I cannot think of anything concrete...
> - "Removed hard failure cases, in favour of finer diagnostics"
> I've NOT merged this one. The globalid variable is used internally.
> It is a fake package identifier that is only used to encode global
> constraints on the universe. It is indeed a fatal error to use this
> globalid otherwise. I cannot see why it should be missing or you
> should have another package to conflict with this id. Maybe you can
> convince me with an example how this could happen ...
> - "Add a check_request function allowing more control over the
> underlying process"
> I merged this one. What kind of finer control do you need on the
> solver sub-process ? Maybe we can also generalize the function
> CudfSolver.execsolver ?
> These two patches are now merged in dose/master.
> Thank you for these patches. Maybe next time, for a faster feedback,
> you could post them directly on the dose development mailing list. I
> track opam-devel, but sometimes I miss one message or an other.
>  dose-devel at lists.gforge.inria.fr
> opam-devel mailing list
> opam-devel at lists.ocaml.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the opam-devel