Brian Duggan on 14 Feb 2019 06:09:28 -0800


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [Philadelphia-pm] Fwd: [perl #133737] Build-time failures with PERL_GLOBAL_STRUCT and PERL_GLOBAL_STRUCT_PRIVATE


Nice!

Thanks for the talk, Jim.

Folks -- next month the second Wednesday doesn't
work for me so I set the date to March 20th --
we can revisit this as it gets closer.

Happy ❤️-day everyone
Brian

On Thursday, February 14, James E Keenan wrote: 
> Coincidentally, today there was some discussion on p5p about two of the
> weird configuration options for Perl 5!
> 
> 
> -------- Forwarded Message --------
> Subject: [perl #133737] Build-time failures with PERL_GLOBAL_STRUCT and
> PERL_GLOBAL_STRUCT_PRIVATE
> Date: Wed, 13 Feb 2019 11:04:18 -0800
> From: Jarkko Hietaniemi via RT <perlbug-followup@perl.org>
> Reply-To: perlbug-followup@perl.org
> To: jkeenan@pobox.com
> 
> [Karl found me gathering dust in a closet]
> 
> So yes, the global struct config was created for Symbian, which is dead as a
> dodo.  The probability of still existing Symbian phones getting Perl
> installed is extremely low.
> 
> But more generally, the config was not Symbian-specific as such.
> It was  created to cover for a limitation in Symbian, and that was
> "shared  libraries shall not have writeable data" (they could have
> data, but only read-only).  In more modern operating systems the
> writeable data sections of shared libraries are copy-on-write
> 
> The global struct is a bit misleading as terms go: what it does is that it
> pulls all the global data into a single struct, which can then be
> heap-allocated in main, and just passed around.  So it is "a struct for
> globals", not a "global" "struct".  The feature comes in two flavors,
> vanilla and PRIVATE,the latter of which is really strict: it leaves
> not even the global struct visible, but instead a function call
> through which you can access the the struct.  (The description
> I wrote (I think) for perlguts is still good.)
> 
> I still find the config useful as a cleanliness exercise: do we have a good
> enough understanding and tracking of the global data so that we
> can move all of it to heap, making the shared library as "pure" as possible.
> But then  again, as the creator of the feature I am biased.
> _______________________________________________
> Philadelphia-pm mailing list
> Philadelphia-pm@pm.org
> https://mail.pm.org/mailman/listinfo/philadelphia-pm
_______________________________________________
Philadelphia-pm mailing list
Philadelphia-pm@pm.org
https://mail.pm.org/mailman/listinfo/philadelphia-pm