John Cremona on Fri, 16 Jul 2010 21:34:25 +0200

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

Re: some glitches in make test-all

[Sorry this is not a proper reply to the thread I started.  I did not
get Karim's reply to my post as an email, only from the online
archive, which does not allow replying.]

Thanks for the detailed reply.

(1) You are right that I had the defaults set for the install
directory, but had not yet done "make install".  I definitely agree
that it should be possible to let users do "make test-all" *before*
"make install".  That should be possible to manage by some use of
environment variables and the like, but I am not volunteering.

(2) I also think it would be excellent to disable the parts of
test-all which require these separate data files if they are not
present.  People may not need my elliptic curve database (there must
be some out there!) but they should be able to run a full test without
seeing "BUG".

I ran into this while working on the integration of 2.4.3 into Sage,
where people like to have the option of running each constituent's
test suites.  In this situation the problem of (1) does not arise
(since Sage's installation does install everything before any check
would be done) but (2) does, since Sage uses galdata and seadata but
not elldata (since it manages its own elliptic curve database, and not
even I need two copies).

By the way, I think it is excellent that pari's test suite is now so
much more thorough than th old "make bench" which I have known for
years;  and it is particularly commendable that after bug-fixes, new
tests are added -- I noticed!