Re: bjam header dependency check broken (and so are the regression test reports)

From: Robert Ramey (Ramey_at_[hidden])
Date: 2005-01-15 14:54:52

For what its worth I've noticed the same problem here. At first I thought it
was limited to *.ipp files but after I started looking at it I could see its
deeper than that. I just presumed that once it seemed necessary to rebuild
all the tests, that everyone knew about this. I had thought is was
considered low priority as v2 would be coming out "real soon now". BTW this
costs a huge amount of time. For example, there have been no changes
checked into CVS since its release - yet the tests are re-run from scratch
periodically.

Robert Ramey

David Abrahams wrote:
> John Maddock wrote:
>> I'm currently having a lot of problems with the regex regression test
>> reports
>>
(http://www.meta-comm.com/engineering/boost-regression/developer/regex.html)
>> that appear to be caused by bjam not noticing when a dependent
>> header has changed (The problems do not appear if the tester does a
>> clean rebuild each time).
>>
>> For example if I cd into libs/regex/example, and
>>
>> bjam credit_card_example
>>
>> to bring that target up to date, and then:
>>
>> touch ../../../boost/regex/v4/cpp*.hpp
>> bjam credit_card_example
>>
>> then nothing gets rebuild (everything should be rebuilt!).
>>
>> However as soon as I touch a source file:
>>
>> touch ../src/cpp*.cpp
>> bjam credit_card_example
>>
>> Then the touched source files do at least get rebuilt.
>
> John,
>
> I've seen other reports of this problem recently. Since you have a
> reproducible case, I can only request that you try to narrow down when
> the problem was introduced. A log(N) binary search over the state of
> boost/tools/build is tedious, but I can't think of another way to
> narrow down the source of the problem.