To comment on the following update, log in, then open the issue:
http://www.openoffice.org/issues/show_bug.cgi?id=84965





------- Additional comments from [EMAIL PROTECTED] Sat Apr  5 06:17:19 +0000 
2008 -------
I'm having a similar problem on LinuxFromScratch with gcc 4.1.2 and the system
STL.  The compiler fails configure's tests, removing -fvisibility-inlines-hidden
results in the following one-line change in the assembler output:

--- conftest.s  2008-04-05 00:32:07.000000000 -0500
+++ conftest.s.pass     2008-04-05 00:30:53.000000000 -0500
@@ -46,7 +46,7 @@
        movq    %rbx, %rsi
        movq    %rsp, %rdi
 .LEHB1:
-       call   
_ZNSt19basic_istringstreamIcSt11char_traitsIcESaIcEEC1ERKSsSt13_Ios_Openmode
+       call   
[EMAIL PROTECTED]
 .LEHE1:
 .L7:
 .L8:

Apparently, on amd64, calls within a shared object MUST be indirected through
the PLT.  Attempting to avoid doing so results in a PC-relative relocation being
required, but amd64 forbids the use of such relocations in shared objects
(position independence is enforced, absolutely).

Using STLport might get around this, but only at the cost of *duplicating the
STL code in each module that needs it* (which surely wastes more space than the
long symbol names in the PLT).

Really, GCC should either give an error or just ignore
-fvisibility-inlines-hidden in the case of building a shared object on amd64. 
It doesn't--GCC appears to trust that you know what you're doing, even when
you're asking for code to be compiled in a way such that linking will be 
impossible.

I'm not sure if this issue also extends to -fvisibility=hidden in the general
case or not (I hope not--the control it provides is over exported symbols sounds
very nice).


---------------------------------------------------------------------
Please do not reply to this automatically generated notification from
Issue Tracker. Please log onto the website and enter your comments.
http://qa.openoffice.org/issue_handling/project_issues.html#notification

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to