[ptx] Hugin - compilation problems

douglas wilkins dgswilkins at yahoo.co.uk
Thu Jan 13 07:07:04 GMT 2005


 --- Steve Wesemeyer <swesemeyer at ntlworld.com> wrote: 
> On Wednesday 12 January 2005 08:23, douglas wilkins
> wrote:

[snip]

> Just downloaded the latest cvs snapshot and managed
> to compiled it.
> However, I had to explicitly tell configure where
> the fftw libraries were 
> installed whereas the previous snapshot ./configure
> still finds them, eg

[snip]

> I traced it to the ax_fftw macro (which is probably
> one of the ones you 
> changed) and I saw that there is now an explicit
> check for amd64 in there and 
> it then looks for fftw in /usr/lib64. However, when
> I did the 
> normal ./configure;make;make install for fftw on my
> AMD64 they were placed 
> in /usr/lib by default. I guess I should have
> specified a different target 
> directory...

Unfortunately, yes...
amd64 is a 'bi-arch' platform. That is to say, 32 bit
libraries live in /path/to/dir/lib while the 64 bit
versions live in /path/to/dir/lib64
Thus, you need to specify the options like this:

./configure --enable-shared --enable-threads
--prefix=/path/to/dir --libdir=/path/to/dir/lib64

when you run configure.
And just for the record, when I compiled wxWidgets I
had to use something like:

LDFLAGS=-L/usr/X11R6/lib64 ../configure --with-gtk
--prefix=/path/to/dir --libdir=/path/to/dir/lib64

> 
> Boost libs provide another slight oddity. In the
> previous CVS version I 
> compiled the actual Boost_Libs path is wrong, it
> should be /usr/lib64 but it 
> didn't affect the compilation.The new snapshot looks
> as if it didn't find 
> them at all but obviously does. Again I looked at
> the diffs and the new code 
> checks for the stuff in the right places (AFAICT)
> but then sets the variables 
> to "" so they don't show up in the output of the
> config. 

One of the changes I made was to make sure that paths
that gcc uses by default are not specified on the
command line. If you want to know what they are, run:

gcc -print-search-dirs

By the way, only the boost headers are used in the
hugin build. The libraries are only needed by the
programs in ./src/tests

> Not really a problem 
> as the compilation still works fine but it got me a
> bit worried especially 
> after finding out that fftw wasn't detected anymore.

see above :-)

> I haven't had any problems with checkinstall so far
> and the previous versions 
> of hugin that I compiled (around summer time last
> year) used to work with it 
> as well. I'm sure once the build process has settled
> across the various 
> architectures that hugin runs on, it might work
> again. But I'll have a look 
> at building an rpm by hand as well...

Maybe you could send me the log of a failed
checkinstall run?

regards,
Doug


	
	
		
___________________________________________________________ 
ALL-NEW Yahoo! Messenger - all new features - even more fun! http://uk.messenger.yahoo.com


More information about the ptX mailing list