From: Beman Dawes (beman_at_[hidden])
Date: 2000-08-17 08:37:04
John Maddock wrote:
>If you download the latest cvs source then that should be fixed
>(hopefully), this brings me on to another point - testing with compilers
>that the library author doesn't necessarily have access to, thoughts that
>spring to mind include:
>
>1) set up a "boost compile farm" in a manner similar to sourceforge,
>probably this requires more infrastructure support that most boost
members
>are able to give.
Yes, a compile farm would be nice. Jens Maurer also suggested it to me in
private email awhile ago.
DSL service is supposed to become available here in six months, although I
live a bit far from the central office. If all the if's get worked out, I
will do something here then. I'm hanging onto an otherwise unused Pentium
Pro 200 Mhz system for that experiment. But don't get your hopes up.
The other compile farm I would like to see is for commercial UNIX systems
like Sun, SGI, and HP. Volunteers?
>2) automate the compile tests (this in part was why I asked about them
>being available via cvs), a shell script something like:
>
>cvs update
>run_test_script
>cvs commit test_result_files
>
>run nightly would be a useful addition. Again though it requires that
>someone with a permanent net connection to be kind enough to set this up!
>Likewise although testing the main trunk of the source is useful, often
it
>would be the branches (containing tentative fixes) that require testing -
I
>don't see any easy way of dealing with this - it would be possible to cvs
a
>"control file" containing a list of branches that authors have requested
be
>tested - but in all likelyhood the test requests could grow unfeasably
long
>unless some strict management is applied.
>
>Thoughts?
We are definitely moving in that general direction.
The next step is a real build system. We have to make it easy for
developers to build and test, and we have to make it easy for users to
download, install, build and if they wish, test.
--Beman