From ff6b7ba98e8d4aab04cbe2bfdffdfc9171c1812b Mon Sep 17 00:00:00 2001 From: Mark Murray Date: Wed, 9 Sep 1998 07:00:04 +0000 Subject: Initial import of Perl5. The king is dead; long live the king! --- contrib/perl5/ext/Thread/Notes | 13 +++++++++++++ 1 file changed, 13 insertions(+) create mode 100644 contrib/perl5/ext/Thread/Notes (limited to 'contrib/perl5/ext/Thread/Notes') diff --git a/contrib/perl5/ext/Thread/Notes b/contrib/perl5/ext/Thread/Notes new file mode 100644 index 000000000000..1505877ee9d4 --- /dev/null +++ b/contrib/perl5/ext/Thread/Notes @@ -0,0 +1,13 @@ +Should cvcache be per CV (keyed by thread) or per thread (keyed by CV)? + +Maybe ought to protect all SVs by a mutex for SvREFCNT_{dec,inc}, +upgrades and so on. Then use SvMUTEX instead of CvMUTEX for CVs. +On the other hand, people shouldn't expect concurrent operations +on non-lexicals to be safe anyway. + +Probably don't need to bother keeping track of CvOWNER on clones. + +Either @_ needs to be made lexical or other arrangments need to be +made so that some globs (or just *_) are per-thread. + +tokenbuf and buf probably ought to be global protected by a global lock. -- cgit v1.3