Re: [Boost-build] Proposal to extend gcc's flavor feature

From: Vladimir Prus (ghost_at_[hidden])
Date: 2008-04-22 03:03:16

On Monday 21 April 2008 21:12:56 Roland Schwarz wrote:
> Currently it is hard to condition on a specific
> incarnation of gcc.
> 
> While it is possible to condition on a certain
> version of gcc, this still is unreliable, since
> the version is, what the user specifies in
> user-config.jam.
> 
> A special subfeature named <flavor> exists, mainly
> to disambiguate between mingw and "normal" gcc's.
> 
> The flavor still has it's shortcoming though, since
> if not explicitly requested, it will get the value
> of the first configured gcc. This most likely will
> come to a surprise of the user.

You mean, even you you request second configured gcc explicitly,
you'll get the flavour of the first? That's a bug, and I think
the part of your patch that addresses that is OK.

> Since the flavor is just a special case of
> -dumpmachine I suggest making flavor more general,
> i.e. make flavor the string that is returned by
> -dumpmachine for all gcc's.
> 
> Benefits:
>   The incarnation of gcc can be uniquely referenced,
>   which is of particular importance in cross builds.
>   (Where mingw/cygwin is only a special case.)
> 
>   It is mostly compatible, i.e. the lib tagging
>   continues to work as current.
> 
> Drawbacks:
>   The directory names are getting longer, since flavor
>   is a non-incidental feature.
> 
> 
> Is there a way to make the feature assymetric, so that
> the default gcc would not suffer from the longer names?
> 
> I attach a diff to current gcc.jam to this mail.
> 
> I am interested to hear what you think about this proposal.

1. What additional information will be available that is not
available now? I presume, the real version?

2. Why should the flavour be present in the path at all?

- Volodya

> 
> Regards,
> Roland
> 
>