John Cremona on Wed, 21 Jul 2010 23:10:51 +0200

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: some glitches in make test-all

On 21 July 2010 21:55, Bill Allombert
<> wrote:
> On Wed, Jul 21, 2010 at 12:04:59PM +0100, John Cremona wrote:
>> I was able to fix my specific problem (opting not to test
>> ellglobalred) by changing line 7 of config/get_tests from
>> test_extra_out="ploth"
>> to
>> test_extra_out="ploth ellglobalred"
>> Now, it should not be hard to make this automatic by testing the
>> existence of the data directory $GP_DATA_DIR/elldata and similarly to
>> include ell-sea in that list iff  $GP_DATA_DIR/seadata does not exist.
>>  By scripting skills are not so good, and I'm sure the person who
>> wrote get_tests could do that faster and more reliably than me!
> Well, the purpose of 'make test-all' is to test everything (as long as we have
> written tests), so this requires potentially that all the optional packages are
> installed.  Making a commitment that 'make test-all' will work with
> such-and-such subset of the packages would not make sense.  So would blindly
> omiting tests, because the user would assume the tests did pass while they were
> not performed at all.

I disagree!     A user who does not need the elliptic curve database
will clearly not install it;  but he is then told that there is a bug
in ellglobalred, which is misleading.  I am just suggesting that
make-all acts like make-all-installed.

> What you actually want is another 'make test-*' target that do what you want.

Yes: make-all-installed.

Or: for a test that fails only because of a not-installed data file,
the output could read something other than "BUG" (and the result code
should perhaps be OK).


> Cheers,
> Bill.