Summary: | app-text/sary: libsary is underlinked, cause failures on dev-ruby/sary-ruby | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Jeroen Roovers (RETIRED) <jer> |
Component: | [OLD] Library | Assignee: | CJK Team <cjk> |
Status: | RESOLVED FIXED | ||
Severity: | normal | Keywords: | PATCH |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 372079, 457238 | ||
Attachments: |
dev-ruby:sary-ruby-1.2.0-r3:20120527-145513.log
remove the superfluous have_* checks dev-ruby/sary-ruby-1.2.0-r3/work/ruby18/sary-ruby-1.2.0/mkmf.log |
Created attachment 313297 [details, diff]
remove the superfluous have_* checks
Could you also post /var/tmp/portage/dev-ruby/sary-ruby-1.2.0-r3/work/ruby18/sary-ruby-1.2.0/mkmf.log ? I noticed -lsary check is only failed with ruby18. We need mkmf.log to determine the problem.
> [32;01m*[0m Running configure phase for ruby18 ...
> checking for sary.h... yes
> checking for main() in -lsary... no
> [32;01m*[0m Running configure phase for ruby19 ...
> checking for sary.h... yes
> checking for main() in -lsary... yes
Created attachment 317630 [details]
dev-ruby/sary-ruby-1.2.0-r3/work/ruby18/sary-ruby-1.2.0/mkmf.log
I can't reproduce this after rebuild app-text/sary. Oh, it's back. On any arch I get this: $ scanelf -n /usr/lib/libsary.so TYPE NEEDED FILE ET_DYN libc.so.6 /usr/lib/libsary.so whereas libsary does need a lot of symbols from dev-libs/glib: # scanelf -s'*' /usr/lib/libsary.so|grep g_ (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_malloc (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_free (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_hash_table_lookup (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_assertion_message_expr (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_array_new (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_strerror (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_strconcat (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_hash_table_new (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_array_free (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_array_append_vals (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_malloc_n (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_hash_table_insert (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_log (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_hash_table_destroy (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_strdup (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_assertion_message (libsary.so) 0 STT_NOTYPE STB_GLOBAL SHN_UNDEF g_hash_table_foreach I guess the real problem is in app-text/sary. Somehow the HPPA linker is bothered by that, whereas the x86/amd64 linkers are not. Some versions of the x86/amd64 linkers will be bothered. This is an underlinking problem and, for instance, it won't sit well with the gold linker. So neither arch-specific nor toolchain-related. Fixed in -r1. |
Created attachment 313295 [details] dev-ruby:sary-ruby-1.2.0-r3:20120527-145513.log have_library('sary') fails on HPPA, probably because of some pthreads weirdness I don't quite begin to understand. extconf.rb actually does a fine job of employing pkg-config to figure out whether sary is installed and how to compile and link against it, so the have_* checks should probably go away.