[Zlib-devel] CHOST option

[Volker Grabsch]

Enrico Weigelt <weigelt at metux.de> schrieb:
> * Volker Grabsch <vog at notjusthosting.com> schrieb:
> 
> > > This only works in the specific case that you have only *one*
> > > toolchain per target-tuple and its commands are available
> > > under these names.
> > 
> > That is by far the most common case, so providing a handy
> > shortcut for that case is a very sensible thing to do.
> 
> Maybe in number of performed builds, but not technically.
> BTW: that approach *only* works if you have only one toolchain
> per target-tuple, otherwise you'll easily get into big trouble.

I must confess that I don't understand you critism. I never
questioned that absolute paths are sometimes needed. I even
showed that both approaches don't exclude each other:

> > However, in that case the first approach ("make CROSS=i686-pc-mingw32-")
> > would still work, wouldn't it?
> > 
> >     make CROSS=/path/to/toolchain/bin/
> > or:
> >     make CROSS=/path/to/toolchain/bin/i686-xx-yyy-
> > 
> > And in the end, zlib's CROSS_PREFIX is exactly that, isn't it?
> 
> I'll first have to check it carefully to answer this. My build 
> machinery passes individual toolchain commands - for almost a
> decade now - as this is the de facto standard. (for those who
> don't, I mainain my own branch).

I fully agree that AR, CC, etc. should override the defaults
derived from CROSS (or CHOST). If that's not the case, it is
really a bug in zlib.


Greets,

    Volker

-- 
Volker Grabsch
---<<(())>>---
Administrator
NotJustHosting GbR