[Xapian-discuss] Threaded test (in C++) to reproduce our database problems

James Aylett james-xapian@tartarus.org
Thu, 17 Jun 2004 11:56:57 +0100


On Thu, Jun 17, 2004 at 11:24:16AM +0100, Olly Betts wrote:

> > I've got this running happily here on fairly recent CVS code. What
> > OS/processor are you using?
> 
> Somewhere way upthread, Eric said:
> | On my dual-processor Mac running OS X, this test fails every time.

Ah, mea culpa.

> > What compiler?
> 
> I don't think he said.

Probably gcc of some breed.

> > (That code won't compile straight off with g++ 3.0 on Deibha/ia32
> Should Deibha be Debian?

Yeah - I'm using a zero-response keyboard, so some things come out
weird :-)

> > because the type of the pthread target function is wrong; I fixed the
> > code to compile and am not seeing the issue.)
> 
> Is the fix obvious?  I should try this with 0.8.0 in case it has been
> fixed incidentally by some change since.

pthread on Debian wants the target function to (a) take (void*) and
(b) return (void*); Eric's target took (char*) and returned (void), so
there are some casts, a signature change, and a new return
statement. OS X compilation may be ignoring things like this, but I
doubt it'll be causing problems (although it's not healthy).

J

-- 
/--------------------------------------------------------------------------\
  James Aylett                                                  xapian.org
  james@tartarus.org                               uncertaintydivision.org