Changes between Version 1 and Version 2 of dev-lang/perl/5.8.8/1

Timestamp:

01/25/09 10:49:17 (6 years ago)

Author:

brecht

Comment:

--

dev-lang/perl/5.8.8/1

@@ -1 @@
- 
+Following the instructions in the supplied README.beos, it is easy enough to build Perl on BeOS.
+
+On BONE, a large number of tests fail with an error similar to this one:
+
+{{{
+load_add_on(/boot/home/Documents/code/perl-5.8.8/t/../lib/auto/PerlIO/encoding/encoding.so) : -1 (General OS error)
+Can't load '../lib/auto/PerlIO/encoding/encoding.so' for module PerlIO::encoding: General OS error at ../lib/XSLoader.pm line 70.
+ at ../lib/PerlIO/encoding.pm line 14
+}}}
+
+When kernel debugging is enabled, the following is printed to the syslog:
+
+{{{
+KERN 'perl'[849]: clone_area_gen error
+}}}
+
+It seems that Firefox suffered from a similar problem at one point: http://www.bebits.com/talkback/2715?page=14. This was because of the global 32MB add-on limit in BeOS.
+
+Dynamically loading {{{encoding.so}}} in a small test application is not a problem, so that library is not to blame. When linking this small test application with libflock however, the same error occurs - '''sometimes'''! The error doesn't seem to be reproducible but random. Perhaps this has something to do with the 32MB global add-on limit in BeOS?
+
+The author of flock_server, Ingo Weinhold, only develops on R5. Hence, there could be a problem with flock_server on BONE.
+
+When compiling Perl without flock() support (removing flock.h), there are no load_add_on() errors. This does indeed support the idea that there is a problem with flock_server on BONE.
+
+When statically linking all extensions, load_add_on() is not used, and thus there is no errors. This results in a perl executable of 6MB though.
+
+To build Perl with dynamically loaded extensions, we need to find out what the incompatibility between flock_server and BONE is (ticket #38).
+
+The built perl executable will sometimes output the following warning:
+
+{{{
+Can't ignore signal CHLD, forcing to default.
+}}}
+
+This has something to do with the SIGCHLD POSIX signal. To me, it's not entirely clear whether the perl source code is incorrect, or BeOS (ticket #40).