[petsc-dev] bullshit in Plex example under complex

[Barry Smith]

> On Aug 29, 2015, at 6:13 PM, Tobin Isaac <tisaac at ices.utexas.edu> wrote:
> 
> On Sat, Aug 29, 2015 at 05:53:22PM -0500, Barry Smith wrote:
>> 
>>> On Aug 29, 2015, at 5:47 PM, Jed Brown <jed at jedbrown.org> wrote:
>>> 
>>> Tobin Isaac <tisaac at ices.utexas.edu> writes:
>>>> How am I supposed to add a test to the intersection of a package
>>>> (testexamples_TRIANGLE) and a scalar type (testexamples_C_NoComplex).
>>>> It's pretty frustrating, given how obvious this would be with
>>>> something like sharness [1], that I couldn't easily figure this out.
>>> 
>>> I've used sharness a reasonable amount and generally like it, especially
>>> when paired with "prove".  A couple years ago, I did some exploratory
>>> work to parse PETSc makefile tests and generate sharness-style tests.  I
>>> still don't think it's a bad idea and might yet be able to get back to
>> 
>>  It seems to require a lot of boilerplate for each test. I sure as heck hope we don't have to write that by hand for each test.
> 
> Really? Compare sharness in hpgmpg [1] to PETSc, keeping in mind that
> the .out files in PETSc-parlance are incorporated into the test
> script.  Every test in PETSc (a) writes out the command line in full,
> (b) diff's against the expected output, (c) writes out the "Possible
> problem" message, and (d) removes the temporary files.  If sharness
> can meet PETSc's portability needs, that would strike me as a good
> reduction in boiler plate.

  I never said the current system doesn't have boilerplate! It absolutely does, all I am saying is that a replacement should not have any. A replacement has to be fundamentally better than the current model, not just a bit better. For example the MORONs who switch from GNU autoconf to cmake (out of the frying pan into the fire).

  Barry

> 
>  Toby
> 
> [1] https://bitbucket.org/hpgmg/hpgmg/src/26f74a2120cefc078b9b855297b77bf8d62ffbc0/finite-element/test/t010-grid.sh?at=master
> 
>> 
>>  I'm not a fan of the "parse the PETSc makefiles approach"; I think we need to define a standard format for declaring tests that is very concise (only exactly what is needed; I have one try in a branch somewhere). Then we can have code that translates from the standard format to anything people want to try with testing scripts. I am even willing to manually convert makefile crap to this format when necessary 
>> 
>>  Barry
>> 
>>> it.
>>