[Bioperl-l] environment variables (et al.) for bioperl-run tests

[Chris Fields]

On Jan 6, 2009, at 11:44 AM, Amir Karger wrote:

>> Amir Karger wrote:
>>> We just installed Bioperl-run on a cluster with very limited
>> environment
>>> variables. So a huge number of tests failed.
>>
>> None of the tests should fail. It's a bug if they do, so please let
>> us
>> know the details. When an environment variable isn't set or a
>> program
>> not installed, the test should skip, not fail.
>
> Oops! I lied. I should have more clearly stated that most of the tests
> nicely skipped, and some even told me that a certain environment
> variable should be set. The only things that actually broke are:
>
> t/AnalysisFactory_soap....10/12 proxy: transport protocol not  
> specified
> # Looks like you planned 12 tests but only ran 11.
> # Looks like your test died just after 11.
> t/AnalysisFactory_soap.... Dubious, test returned 255 (wstat 65280,
> 0xff00)
> Failed 1/12 subtests
>
> Also, I got an error from RepeatMasker and Vista, which should  
> probably
> be more graceful when things aren't in the PATH:
>
> t/RepeatMasker............Use of uninitialized value in concatenation
> (.) or string at
> /odyssey/apps/src/bioperl-run-1.5.2_100/blib/lib/Bio/Tools/Run/ 
> RepeatMas
> ker.pm line 175.
> RepeatMasker program not found as  or not executable.
> RepeatMasker program not found. Skipping tests 1 to 10.
>
> t/Vista...................readline() on closed filehandle PIPE at
> t/Vista.t line 39.
> Use of uninitialized value in numeric lt (<) at t/Vista.t line 53.
> You need at least version 1.2 of JDK to run vista
>
> It looks like the bug Chris referred me to is a bug to fix the above
> errors, which is great. But if all the tests pass and I haven't  
> actually
> tested any of my bio binaries, I'm going to be kind of bummed when I
> actually try to use them. I'm not sure how flexible the testing
> mechanism is: is there an easy way to give the user a one-line  
> feedback
> when they run the test that says "The following binaries weren't found
> or did not have relevant environment variables set, so their subtests
> were skipped: blastall, clustalw, primer3, ...". That would be  
> awesome.
> Even better would be if it referenced a file to look at or a wiki page
> listing how to set up PATH and other environment variables for each
> binary.
>
> Sorry for the mis-type.
>
> -Amir

I'll look into the test fails if I can reproduce them (they'll be  
archived with the bug report), and add that last bit as an enhancement  
request to bugzilla.

chris