[Boost-testing] Re: Whitespace (again)

From: David Abrahams (dave_at_[hidden])
Date: 2005-03-12 19:19:56

"Victor A. Wagner Jr." <vawjr_at_[hidden]> writes:

> At Saturday 2005-03-12 15:40, you wrote:
>>Victor A. Wagner Jr. writes:
>> > At Saturday 2005-03-12 10:19, you wrote:
>> >>"Victor A. Wagner Jr." <vawjr_at_[hidden]> writes:
>> >>
>> >> > At Friday 2005-03-11 19:09, you wrote:
>> >> >>"Victor A. Wagner Jr." <vawjr_at_[hidden]> writes:
>> >> >>
>> >> >> >
>> >> >> > If I remember correctly, I put the del *.xml into the script because
>> >> >> > if I didn't, and the tests blew up spectacularly, the xml wouldn't be
>> >> >> > deleted, the old one would still be there and nobody would be the
>> >> >> > wiser until someone goes "how come Rudbek's ruing blah test just
>> >> >> > fine?" then I panic and discover that indeed, that some test is
>> >> >> > crashing so hard that the test report is inaccurate.
>> >> >> > I'd prefer _not_ to have to build every test every
>> time...the run time
>> >> >> > goes up appreciably.
>> >> >> > so... what's the solution?
>> >> >> > to quit deleting .xml files and hope the original problem is fixed?
>> >> >> > delete all the .exe only?
>> >> >> > what are .test files?  _should_ I be deleting them??
>> >> >> > "Enquiring minds want to know"
>> >> >> >
>> >> >>
>> >> >>Victor,
>> >> >>
>> >> >>
>> >> >>Due to many reasons incremental testing is _not_ reliable and there is
>> >> >>no quick way to fix that. The list of problems we have with
>> >> >>incremental testing is on http://tinyurl.com/5mk7l. Any help would be
>> >> >>appreciated. All four of the issues require a more than introductory
>> >> >>knowledge of BBv1 - for sure certainly more than I have.
>> >> >>
>> >> >>It would be nice if somebody would take on some of the issues, because
>> >> >>the learning curve for me would be the same as for any bjam novice.
>> >> >
>> >> > so how do I do regression runs that ARE reliable?
>> >>
>> >>If you want to be absolutely sure, pass the -a option to bjam and it
>> >>will treat everything as though it's outdated.
>> >
>> > yeah, but regression.py is handling the calls to bjam not me.
>>
>>'regression.py' allows you to pass options to bjam through
>>conveniently named "--bjam-options" option. However, I'm not sure I
>>understand Dave's suggestion. Is passing "-a" any different than
>>deleting all the binaries -- e.g. doing "full" run?
>
> I don't know, but I've uncommented the del *.obj /s and the del *.exe /s
>
> we'll see how that works

It won't work, as I implied by saying (it's not just binaries, but all
of the build products, some of which are text files).

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com