Perhaps two levels of tagging could be used, one to indicate that the tools
successfully built together and another to indicate that they completely
passed the
regression tests. A tool like mozilla's tinderbox--which constantly
rebuilds a checkout from cvs to test for broken builds--could satisfy the
first objective. Fully regression tested distributions will happen
infrequently,
I suppose. Making the tools available in a form where one can test the
tools,
no the build procedure, seems a worthwhile goal.
At our company we like automatic build checking to screen out simple errors
and ensure that the makefiles are kept in order. Build busting
committers experience
shame and the derision of their peers (well, the derision of one's peers
is pretty
much a given no matter how the build turns out). Fame and fortune may be
good
long term motivators, but in the short run we rely on guilt.
Best Wishes,
Greg Wright
Gregory Wright
PacketStorm Communications
Alastair David Reid wrote:
>Gregory Wright <[EMAIL PROTECTED]> writes:
>
>>Is there interest in tagging the projects in the cvs that
>>build cleanly as a suite?
>>
>
>There's talk of doing exactly that with the new library hierarchy.
>The idea would be to insert a tag every time the whole lot passes the
>regression tests for all compilers they'e supposed to work on.
>Extending this to cover tools would probably be straightforward (in
>fact, it's probably essential since the libraries need ghc, hugs, nhc,
>hsc2hs, greencard, ...). The main doubt is whether we can make the
>regression tests good enough to rely on a purely automatic system.
>
_______________________________________________
Glasgow-haskell-bugs mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
