General cross compilation 'problem'.

[John Clark <jclark at metricsystems dot com>]

For quite a while I've had a pretty negative view on the use of the 
'configure'
set up utility.

What seems to be a great idea appears to be fraught with problems when one
steps outside of the bounds of 'compiling on the target machine' 
problems arise.

For example, when compiling 'bash' for an xscale target, on a x86 host, one
ends up with a number of configure exits with the message 'cannot run 
... while
cross compiling'.

What is stranger is that given the check, given that the configure 
'bombs', why
doesn't someone either 1) insert a 'safe' but possibily stupid default, 
2) check
at the top of the configure, and when a cross compilation is detected, just
plain quit in the first place.

What is worse, is that those packages which are almost 'sure' to be cross
compiled, these 'tests' arise. For example 'dropbear'... if one wanted
ssh setup on a full filled out host, compiler and all, then why bother 
with dropbear...

Anyway, the list goes on of packages that have configure scripts which 
expect to run
something on the 'target', during the configuration process.

My 'method' of dealing with this is to basically had edit the configure 
script
until it completes, eliminating the pattern "{ (exit 1); exit 1; };" in all
cases where the 'if test "$cross_compile" = yes ... starts of  the if 
statement.

John Clark


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