Boost logo

Boost-Build :

From: David Abrahams (dave_at_[hidden])
Date: 2006-05-13 12:51:39


"Felipe Magno de Almeida" <felipe.m.almeida_at_[hidden]> writes:

> Hello,
>
> On 5/11/06, Rene Rivera <grafik.list_at_[hidden]> wrote:
>
> [snipped]
>
>> I'm very doubtful about likely gains in speed. Over time I've looked at
>> the profiling info and current dependency calculation is not the biggest
>> time hog, although it is one of the biggest memory hogs (because of
>> having to scan the sources).
>
> Do you have those info still available?

You can generate them yourself with -d+10.

>> I understand that a content specific solution would yield smaller change
>> sets. But the complexity involved I don't think outweighs the minimal
>> benefit.

Here I think Rene is missing the biggest benefit: bigger change sets
:). Right now our dependency resolution can't see through these

   #include SOME_MACRO(...)

As a result, we miss some dependencies. That causes some regression
runners to do a clean run, for accuracy, when they should do an
incremental run.

In other words, it's not important to make the change sets smaller;
it's important to make them more accurate.

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

Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk