Nicholas Clark wrote:
>==== //depot/perl/regcomp.c#340 (text) ====
>
>@@ -4983,9 +4983,14 @@
> (SvTYPE(new_comppad) == SVt_PVAV) ?
> new_comppad : Null(PAD *)
> );
>+ OP_REFCNT_LOCK;
> if (!OpREFCNT_dec((OP_4tree*)r->data->data[n])) {
>+ OP_REFCNT_UNLOCK;
> op_free((OP_4tree*)r->data->data[n]);
> }
>+ else {
>+ OP_REFCNT_UNLOCK;
>+ }
>
> PAD_RESTORE_LOCAL(old_comppad);
> SvREFCNT_dec((SV*)new_comppad);
>
>
The above seems to be the cause of the problem, since now re.dll doesn't
link:
link -out:..\..\lib\auto\re\re.dll -dll -nologo -nodefaultlib
-debug -opt:ref,icf -libpath:"c:\perl\lib\CORE" -machine:x86
re_exec.obj re_comp.obj re.obj ..\..\lib\CORE\perl59.lib oldnames.lib
kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib
advapi32.lib shell32.lib ole32.lib oleaut32.lib netapi32.lib uuid.lib
ws2_32.lib mpr.lib winmm.lib version.lib odbc32.lib odbccp32.lib
msvcrt.lib -def:re.def
Creating library ..\..\lib\auto\re\re.lib and object
..\..\lib\auto\re\re.exp
re_comp.obj : error LNK2001: unresolved external symbol __imp__PL_op_mutex
..\..\lib\auto\re\re.dll : fatal error LNK1120: 1 unresolved externals
NMAKE : fatal error U1077: 'link' : return code '0x460'
Stop.
Unsuccessful make(re): code=512 at buildext.pl line 146.
NMAKE : fatal error U1077: '..\miniperl.exe' : return code '0x2'
Stop.
Reverting the above regcomp.c change fixes things.
Note that ext/re/re_comp.c is copied from regcomp.c, which didn't cause
any problems when built into the main Perl library. Likewise, similar
changes to pad.c and sv.c didn't cause any problem, so I am guessing
that the use of OP_REFCNT_(UN)LOCK is OK in the core (PERL_CORE?), but
not within extensions?
So it probably comes down to needing to export PL_op_mutex?
My perl59.lib definitely doesn't have the symbol exported, so how do I
make that happen (or would that not be the right thing to do anyway)?
makedef.pl already skips it for non-ithreads builds, but it doesn't seem
to be getting included for ithreads builds anyway :(
Grepping around a bit, I found an old change for a possibly similar
problem (19484) which mentions that "PL_op_mutex [...] isn't exported as
such", but it doesn't elaborate.
- Steve
------------------------------------------------
This email has been scanned for viruses and content by the Radan Computational
Webshield Appliances.
