Re: Final toolchain dir

[Michael Zintakis <michael dot zintakis at googlemail dot com>]

> > I agree with you though - the toolchain should be read only and 
> > preserved clear not least because if there is some kind of error or 
> > fault during the image build and files get overwritten/deleted then the 
> > whole toolchain needs to be built again and that is a bit of a pain!
> >     
>
> Yep! And especially when the toolchain is a ressource shared by many
> developpers, you do not want any of them to fumble in the sysroot.
>   
I've stumbled across a similar scenario today as I was fighting with an 
ancient version of expect (the tcl package), which "does not 
crosscompile" and I had to use some ingenious hacks to get around it. 
When I ran configure, make and then make install the latter failed - 
miserably so - and it turned out that "make install" was attempting to 
write to my toolchain's sysroot (failing miserably as the toolchain is 
obviously read only) - so, if that wasn't the case my toolchain would 
have been compromised - as it turned out the "make install" was a bit 
dodgy! Patching it did the trick though and all worked well in the end.


> The dev-team mostly has no knowledge about what a toolchain is (trust
> me, a basic C++ or Java developper has no idea what a compiler is!
> Especially the Java devel has no clue! Hehe! :-] ).
>   
I know - I am a developer and until about a week or so ago I've only 
used basic "toolchains" (i.e. compiling i686 code on x86_64 machine) and 
that was pretty much my experience of this - until now - when I have to 
build tested ppc and later on arm package to go with it. Not an easy 
task as the learning curve for me is a bit steep!

> BTW gcc-4.6.0 is out!
>   
I'll stick with the tried-and-tested for the time being - I am not one 
who jumps at the newest and greatest straight away - I've burned my 
fingers far too many times to care! :-X


--
For unsubscribe information see http://sourceware.org/lists.html#faq