develooper Front page | perl.perl5.porters | Postings from April 2014

Re: Perl 5.20.0 Blockers, 2014-04-07

Thread Previous | Thread Next
Nicholas Clark
April 8, 2014 10:04
Re: Perl 5.20.0 Blockers, 2014-04-07
Message ID:
On Mon, Apr 07, 2014 at 07:15:12PM -0400, bulk88 wrote:
> Steve Hay wrote:

> > With this patch I have perl building in about 70% of the time it takes
> > without it, and all tests pass. Seems worthwhile to me. I would apply
> > it but for the comment here:
> > 
> >
> > 
> > Should I apply it anyway? -- It seems a shame to miss the other two
> > optimizations in the patch just because there is unfinished business
> > on the 'lib' placement change, and it won't stop more work being done
> > on that in the future anyway... Maybe raise a 5.21.1 blocker ([perl
> > #121332]) to ensure it gets followed-up, not forgotten?
> sounds like a feature request and would probably involve changing EUMM 
> and researching which platforms have atomic directory moves plus 
> possible issues if some .pm are preprocessed/generated by Makefile.PL or 
> other .PLs or something. Since Win32 is the slowest building platform 
> since it doesn't do parallel anything, I'd like to get that patch in. 
> #121119 should be applied, then closed, then Zefram or other maintenance 
> person takes the atomic dir mov idea and create a new ticket for it. The 
>   win32-specific changes can always be undone in the future.

Yes, decouple it from the suggestion in that message.

"It shouldn't be hard" is a significant understatement.
bulk88's assessment of what it would take is a lot more realistic.

It may not even be possible to do reliably, and it certainly would introduce
more complexity into an already complex codebase, which needs to be absolutely
reliable and portable more than it needs to be fast.

Nicholas Clark

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About