Re: [Boost-testing] Toolset naming and regressions

From: Jeff Garland (jeff_at_[hidden])
Date: 2005-06-04 11:57:03

On Sat, 04 Jun 2005 11:33:56 -0500, Rene Rivera wrote
> Jeff Garland wrote:
> > So...we really need to standardize toolset naming for regression testers. 
> 
> It is standardized as: [toolset]-[version]-[platform]-[cpu]-[build]

I don't see that standard being followed b/c not a single toolset I see has
all these elements on the regression pages.  Plus there are are variations
like 'msvc-stlport' -- which I'm not sure how the form above accounts for this
variant.  Also 'vc7' is lacking a dash and has a different name from msvc, etc.  

> As far as the release and Beman's gcc tests. Don't worry about it, 
> i.e. don't do anything about it. Only toolsets marked as required 
> for release are important right now. And *only* toolsets that are 
> properly versioned are marked for release.

I realize I don't have to look at them for release, but this is an ongoing
problem with regression testers coming and going.  Before we've reached this
point I had to rework a bunch of regression markup simply b/c the regression
testers had changed the names of toolsets -- not because of any real state
change in library or compilers.  Also, I'm very interested in making sure gcc
on Solaris continues to work so I need to look at the bigger list (btw, is
there some reason this isn't one of the release compiler set?).  When I look
at the bigger list, it makes me wonder why fft-icc_8.1.30, which appears to be
similar to Martin Wille's intel-8.1-linux, has tons of failures -- so then I
start looking at it spending time...

Jeff