|
|
Boost Testing : |
From: Anthony Williams (anthony_w.geo_at_[hidden])
Date: 2006-11-22 06:19:37
"AlisdairM" <alisdair.meredith_at_[hidden]> writes:
> Anthony Williams wrote:
>
>> I guess it depends on your point of view. I always compile all my
>> projects in some form of "release configuration", with more
>> optimization and less debug info (except for Borland). Anything that
>> doesn't work in that configuration is therefore unusable for me. I
>> would be disappointed if boost shipped a library that claimed support
>> for one of my compilers, but actually failed to work in a release
>> build; I would naturally assume it was my code that was at fault and
>> not the boost library.
>
> I'm not disputing the value or necessity of the testing - I am
> disputing the timing of adding a new test platform, which is
> effectively what you are proposing.
And that's where we differ --- I don't see it as a new test platform. If we
say MSVC 7.1 is supported, and all the tests pass, as a user, I (at least)
would assume that it would work in release builds. If this is not the case,
then the library is broken, from my perspective as a user.
I think we should have been running release mode tests all along. In any case,
it is better to run them now, and find problems before 1.34 gets shipped, so
we stand a chance of fixing them, rather than after release when we have to
wait for 1.35 (and how long will that take?)
I see it as a general issue. Knowing that there are potential tests that could
be run *on compilers we claim to support* do we:
a) Run the tests, and try and fix any problems
b) Run the tests, and say "these are known issues" (expected failures).
c) Don't run the tests.
I'd opt for a), every time. If we really can't fix the problems, I would
choose b) as the fallback. I would never choose c), however close to the
potential release date we are.
> Also, my perspective is very clouded as a Borland user as almost 12
> months after the current product was launched there is still no
> official Boost support - yes, Borland managed to time their release to
> be just a couple of weeks after the last Boost cycle, AGAIN! Until
> 1.34 ships, there is no 'official' support for any Borland compiler
> less than 5 years old, so yes my perspective is very skewed ;?)
I agree that we need to get 1.34 out the door ASAP. I just don't want us to
ship a broken library.
>> So, whilst I agree that these failures may have only just come to
>> light, and that it is going to be extra work to fix them, I think it
>> is important that boost works (and therefore is tested) in the
>> configuration(s) of the compilers actually used by boost users.
>
> And that is why I think they should be a high priority for 1.35, and am
> gearing up to work on the Borland issues as we speak!
Thanks.
Anthony
-- Anthony Williams Software Developer Just Software Solutions Ltd http://www.justsoftwaresolutions.co.uk