|
|
Boost Testing : |
From: Vladimir Prus (ghost_at_[hidden])
Date: 2007-05-26 16:27:56
Stefan Seefeld wrote:
> Vladimir Prus wrote:
>> Rene Rivera wrote:
>>
>>> I started doing RC_1_34_0 tests earlier this week, using gcc-4.2.0, and
>>> noticed the following in the logs during the checkout:
>>>
>>> cvs update: Updating libs/graph/src
>>> M libs/graph/src/graphviz_digraph_lex.cpp
>>> M libs/graph/src/graphviz_graph_lex.cpp
>>> cvs update: Updating libs/graph/src/python
>>>
>>> Somehow the process of testing, or building, changes those files. This
>>> doesn't seem like a good thing to me, as it prevents those files from
>>> getting updated from CVS. Or worse if someone changes them in CVS, it
>>> may cause merge conflicts in the update, which would produce invalid
>>> files, breaking whatever tests depend on those.
>>
>> I think those files are generated by flex, and probably Boost.Build
>> thinks those files in CVS are out-of-date with respect to flex source.
>> Maybe checking in the modified version will be sufficient.
>
> What's the reason for having them in the repository then, in the first
> place ? Can't flex be made a precondition for compiling ?
I don't know. In fact, it seems that flex-based parser will
be compiled only if GRAPHIVZ_PARSER variable is set, which
should not set by default.
> How does
> boost.build decide whether it the cpp is up to date ?
Just like it does for all other (target,source) pairs?
> I'm asking also because I have noticed last week that some html is in the
> repository, even though the sources (qbk) it is made from is there, too.
>
> What are the guidelines for what belongs into the repository and what not
> ?
I have no comment on the above two questions.
- Volodya