Re: [Denemo-devel] Master is badly broken

[Richard Shann]

On Wed, 2014-06-25 at 17:01 +0200, Éloi Rivard wrote:
> I don't know. Playing with user's files seems a bit dirty to me.
I understand your concern - imagine some catastrophic failure of the
test harness that caused it to exit after moving the user's /denemo-xxx
away and before restoring it. We should tell the user where it is going
in any case, but it would mean the naive user would be reset to the
first time user experience. Well, naive users don't run tests (but then,
their distro might do it for them).


>  Maybe specify the ".denemo-xx " dir to use with an argument would be
> a good solution, for tests.
That would less invasive than previous proposals, there is only one
place where .denemo-xxx is generated in the code so implies only special
code at that point.

Richard


> 
> 
> 
> 2014-06-25 16:51 GMT+02:00 Richard Shann <address@hidden>:
>         On Wed, 2014-06-25 at 14:19 +0200, Éloi Rivard wrote:
>         > Don't you think the easiest solution would be to pass an
>         argument to
>         > denemo that makes just ignore .deneno-xx files ?
>         
>         
>         As a general principle it is best to create the test harness
>         around the
>         actual executable that the user is going to run, without
>         special
>         parameters being passes (which cause different code to execute
>         than the
>         code the user will ever see). In this case all sorts of
>         consequences
>         could flow from not having the .denemo-xxx directory
>         functioning in the
>         way the program expects it to.
>         I think the ideal way to proceed would be for the test harness
>         to
>         re-name ~/.denemo-xxx to a temporary name (if it exists) and
>         to copy a
>         standard one in its place. At the end of testing
>         the .denemo-xxx is
>         moved into the logs and the user's .denemo-xxx is restored
>         from the
>         temporary location.
>         In this way the program is running in a normal environment
>         which is
>         completely known. Eventually we could have tests that would
>         check for
>         the correct behavior given specific .denemo-xxx files.
>         
>         Richard
>         
>         
>         >
>         >
>         > 2014-06-25 14:09 GMT+02:00 Richard Shann
>         <address@hidden>:
>         >         On Wed, 2014-06-25 at 13:20 +0200, Andreas Schneider
>         wrote:
>         >         > What I meant is a denemo configuration directory
>         included in
>         >         the test
>         >         > suite (i.e. as a subdirectory in the tests
>         directory) with
>         >         standardised
>         >         > values.
>         >
>         >
>         >         hmm, yes, if a user wants to run the tests then
>         having
>         >         the .denemo-xxx
>         >         location fixed at the user's home directory is a
>         problem.
>         >         I think your first suggestion of a command line flag
>         to choose
>         >         where to
>         >         find .denemo-xxx might be good - I have used a
>         separate,
>         >         clean, user for
>         >         this, but it sounds like you need to run the tests
>         as part of
>         >         your
>         >         install process, so that they have to be run as your
>         own user.
>         >         I wonder
>         >         what other projects do?
>         >
>         >         Richard
>         >
>         >         >  Using the user's configuration brings in a random
>         >         component. For
>         >         > example, when I build a package for Denemo (using
>         fakeroot),
>         >         the success
>         >         > of the thumbnailer test depends on which file I
>         had been
>         >         previously
>         >         > working on (in the production environment).
>         >         >
>         >         > Andreas
>         >         >
>         >         >
>         >         > Am 25.06.2014 12:59, schrieb Richard Shann:
>         >         > > On Wed, 2014-06-25 at 10:21 +0200, Andreas
>         Schneider
>         >         wrote:
>         >         > >> Is it possible to set the location of
>         the .denemo-xxx
>         >         directory with a
>         >         > >> command line switch?
>         >         > > There is none at present AFAIK
>         >         > >> I suppose it would be good to include a fixed
>         >         > >> configuration (i.e. a fixed .denemo-xxx
>         directory) in the
>         >         test suite.
>         >         > >
>         >         > > It is at a fixed location relative to the user's
>         home
>         >         directory. The
>         >         > > only problem that I can imagine would be if the
>         test suite
>         >         is run by a
>         >         > > user who has no home directory (or not one that
>         the user
>         >         has permissions
>         >         > > to read and write to). I'm guessing that Eloi's
>         test
>         >         system as run by
>         >         > > travis does not have that problem...
>         >         > >
>         >         > > Richard
>         >         >
>         >         >
>         >         > _______________________________________________
>         >         > Denemo-devel mailing list
>         >         > address@hidden
>         >         >
>         https://lists.gnu.org/mailman/listinfo/denemo-devel
>         >
>         >
>         >
>         >         _______________________________________________
>         >         Denemo-devel mailing list
>         >         address@hidden
>         >         https://lists.gnu.org/mailman/listinfo/denemo-devel
>         >
>         >
>         >
>         >
>         > --
>         > Éloi Rivard - address@hidden
>         >
>         > « On perd plus à être indécis qu'à se tromper. »
>         >
>         
>         
>         
> 
> 
> 
> -- 
> Éloi Rivard - address@hidden
>         
> « On perd plus à être indécis qu'à se tromper. »
>