Re: [jamboost] Re: Some comments on Boost.Jam and Boost.Build

From: Rene Rivera (grafik666_at_[hidden])
Date: 2003-09-01 14:25:17

[2003-09-01] Vladimir Prus wrote:

>Hi Jurko,
>
>> --- In jamboost_at_[hidden], Vladimir Prus <ghost_at_c...> wrote:

>> > > 8. A funny side effect of the help system. If you do
>> > > bjam --help --help
>> > > the help screen is displayed twice :-))
>> >
>> > Funny... but how to fix it. For example, should
>> >
>> > bjam --help doc --help doc
>> >
>> > produce help on "doc" twice, or not ;-)
>>
>> My bet would be on not. The help options already affect all of the
>> help actions - whether triggered by parameters before setting those
>> options or after. So I guess all the help parameteres should be
>> processed first and only then should all the needed actions be
>> triggered. But whether this behaviour can be implemented at all
>> using the current parameter processing system - I can not tell. I
>> can look into it later if you wish.
>
>Let's start by solving the first issue you've raised about help system 
>(--help-all option). I'm waiting for Rene to comment on it, and if he's
busy 
>will dive there myself. 

Did some looking around to figure out what the problem might be. There are
two possibilities: either the alpha order for the list of discovered modules
is incorrect and therefore the do-scan calls are done in the wrong order, or
the "all" target doesn't process it's targets in added order.

For the first one I've put in some changes to options/help.jam to fix it. I
also did some cleanup to reduce the number of bogus discovered modules. So
Jurko if you could try this out that would be wonderfull :-)

For the second, the problem is a bit harder to fix... What the do-scan rule
does is it adds module.scan targets to the all fake target, of which only
the last one has the print(ECHO) "action" (not a real Jam action, just a
rule which does the ECHOs). So if the targets for the "all" are not done in
the added order it could cause the strange behaviour. -- I really don't know
if this is the case though. Fixing it would involve some kind of chaining on
the targets :-(

PS. All the logic for this part is in the doc.do-scan rule, and is fairly
easy to understand :-)

-- grafik - Don't Assume Anything
-- rrivera (at) acm.org - grafik (at) redshift-software.com
-- 102708583 (at) icq