Opened 19 years ago

Closed 18 years ago

Last modified 18 years ago

#1419 closed defect (fixed)

mapserver build break when no include for libiconv could be determined

Reported by: dirk@… Owned by: mapserverbugs
Priority: high Milestone:
Component: Build Problems Version: 4.4
Severity: minor Keywords:
Cc:

Description

The configure script will insert a "-I" without a directory appended when for
some reason no specific include directory could be determined. This happens
because this condition is not tested in configure.in.

The build was done on SuSE Enterprise 9 with Intel 32 bit architecture.
A patch that solved the problem will be attached.

Attachments (1)

mapserver-4.4.2-iconv.diff (592 bytes ) - added by dirk@… 19 years ago.
Patch that fixed my compilation problem

Download all attachments as: .zip

Change History (4)

by dirk@…, 19 years ago

Attachment: mapserver-4.4.2-iconv.diff added

Patch that fixed my compilation problem

comment:1 by dirk@…, 19 years ago

mapserver-4.6.0 should still have this problem. I couldn't see any modifications
in the modified section. However, I didn't test it yet.

comment:2 by dmorissette, 18 years ago

Resolution: fixed
Status: newclosed
Unfortunately I'm not setup to test this configuration, but the patch is
straightforward and (seems to) make sense, so I have applied it in the 4.8 CVS,
it will be in the 4.8 release. I'm actually wondering how the build can complete
if iconv is enabled but iconv.h is not found, but I won't research any further
if you tell me that this combination is valid on some systems.

Would you be able to test the fixed configure script from the CVS HEAD to
confirm that the problem is fixed?

comment:3 by dirk@…, 18 years ago

With CVS-HEAD it compiles. I actually figured out where that issue comes from. I
configured mapserver just with with --with-libiconv ("..." is my root):

./configure --with-gdal=.../bin/gdal-config --with-proj=...
--with-php=.../src/php-4.4.0  --enable-debug --with-libiconv --with-gd=... 
--with-postgis=.../bin/pg_config

My iconv is in libc, which it will be linked against anyways and that's why it
worked. I should have better submitted a patch that catches the missing argument
or tries to determine the locations of iconv.h and libiconv/libc.
Note: See TracTickets for help on using tickets.