another threads reliability fix: serialize writes to thr->threadsv

avoid most uses of PL_na (which is much more inefficient than a
simple local); update docs to suit; PL_na now being thr->Tna may
be a minor compatibility issue for extensions--will require dTHR
outside of XSUBs (those get automatic dTHR)

p4raw-id: //depot/perl@2387

1 files changed, 1 insertions, 1 deletions

diff --git a/perly.y b/perly.y
index 78378b62be..88eee99f3e 100644
--- a/perly.y
+++ b/perly.y
@@ -304,7 +304,7 @@ startformsub:	/* NULL */	/* start a format subroutine scope */
 			{ $$ = start_subparse(TRUE, 0); }
 	;
 
-subname	:	WORD	{ char *name = SvPV(((SVOP*)$1)->op_sv, PL_na);
+subname	:	WORD	{ STRLEN n_a; char *name = SvPV(((SVOP*)$1)->op_sv,n_a);
 			  if (strEQ(name, "BEGIN") || strEQ(name, "END")
 			      || strEQ(name, "INIT"))
 			      CvUNIQUE_on(PL_compcv);