New version of swig


#1

From: http://www.swig.org/

“SWIG-1.3.38 has been released. It contains an output directory
regression fix and other minor bug fixes.”

I have a memory of someone having an issue with swig recently, was it
with gnuradio?

Philip


#2

On Feb 1, 2009, at 8:22 AM, Philip B. wrote:

“SWIG-1.3.38 has been released. It contains an output directory
regression fix and other minor bug fixes.”

I have a memory of someone having an issue with swig recently, was it
with gnuradio?

Your memory is correct, and, yes, the issue was with how SWIG 1.3.37
behaved w/r.t. GNU Radio: in 1.3.36, the default output directory for
all created files was the same as that the argument to the “-o” output
file specifier option. In 1.3.37, the default output directory was
changed for the .py output file to be the directory of the input .i
file … which creates an error when doing ‘make distcheck’ because
the directory of the input .i file is not writable. Adding in the
flag “-outdir $(builddir)” fixes the issue all the way back to around
SWIG version 1.3.20, and should be forwards API compatible as well.
There is a GNU Radio branch which corrects the behavior to make it the
same with 1.3.36 and 1.3.37 (by adding “-outdir $(builddir)”) – which
is still in testing – but 1.3.38 should have corrected this issue for
those who are or will be using it. - MLD


#3

On Sun, Feb 1, 2009 at 12:49 PM, Michael D. removed_email_address@domain.invalid
wrote:

files was the same as that the argument to the “-o” output file specifier
option. In 1.3.37, the default output directory was changed for the .py
output file to be the directory of the input .i file … which creates an
error when doing ‘make distcheck’ because the directory of the input .i file
is not writable. Adding in the flag “-outdir $(builddir)” fixes the issue
all the way back to around SWIG version 1.3.20, and should be forwards API
compatible as well. There is a GNU Radio branch which corrects the behavior
to make it the same with 1.3.36 and 1.3.37 (by adding “-outdir $(builddir)”)
– which is still in testing – but 1.3.38 should have corrected this issue
for those who are or will be using it. - MLD

1.3.38 is working fine in Angstrom for the Beagle :slight_smile:

Still trying to work out my exact issues with make check.

Philip