Boost logo

Boost-Build :

Subject: Re: [Boost-build] Use of alias to not build a toolset is not working
From: Edward Diener (eldiener_at_[hidden])
Date: 2012-11-03 12:18:51


On 10/30/2012 9:46 PM, Steven Watanabe wrote:
> AMDG
>
> On 10/30/2012 05:48 PM, Edward Diener wrote:
>> On 10/29/2012 11:40 PM, Steven Watanabe wrote:
>>> On 10/29/2012 07:50 PM, Edward Diener wrote:
>>>> On 10/28/2012 11:34 PM, Steven Watanabe wrote:
>>>>>
>>>>> Try <toolset>gcc <toolset-gcc:version>3.4.5.
>>>>> (I believe that <toolset>gcc-3.4.5 is not
>>>>> expanded correctly in the context of target
>>>>> alternative selection)
>>>>
>>>> Your syntax worked. Is it a bug that <toolset>gcc-3.4.5 does not work
>>>> for target alterantive selection, while it works in other contexts ?
>>>>
>>>
>>> Yes, I consider it a bug. It is, however, a bit
>>> tricky to handle correctly, since we have to avoid
>>> causing an error when there is no gcc-3.4.5 toolset.
>>
>> Do you mean when the user invokes bjam without specifying
>> toolset=gcc-3.4.5 ?
>>
>
> No. I mean that in order for Boost.Build to
> know that 3.4.5 is the version (and not some
> other subfeature), either there has to be a
>
> using gcc : 3.4.5 ;
>
> or the version deduced by running gcc --version has
> to be 3.4.5. The point is that target alternatives
> whose requirements include unknown subfeature values
> have to be ignored.

If you see <toolset>gcc-3.4.5 in an alias such as:

alias ttivm16 : :
         <toolset>gcc-3.4.5
     ;

why is it not enough to know that if the invocation is:

bjam toolset=gcc-3.4.5 ttivm16

nothing is built without having to know whether gcc --version is going
to be 3.4.5 ? Even if you decide to require that gcc --version still
produce 3.4.5, which I surmise is normally the case when bjam is invoked
with that toolset, I still do not understand the problem as opposed to
the successful <toolset>gcc <toolset-gcc:version>3.4.5 .

I will take your word for it that the problem exists in the Boost build
system but logically it seems it should not.


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