Carview!

CARVIEW

MOTORHOMES

Select Language

HTTP/2 200 cache-control: max-age=43200 server: Combust/Plack (Perl) vary: Accept-Encoding content-encoding: gzip content-length: 3270 content-type: text/html; charset=utf-8 last-modified: Sat, 11 Oct 2025 21:48:26 GMT date: Sat, 11 Oct 2025 21:48:26 GMT strict-transport-security: max-age=15768000 Re: my $@ - a proposal - nntp.perl.org

Front page | perl.perl5.porters | Postings from June 2008

Re: my $@ - a proposal

Thread Previous | Thread Next

From:

Yuval Kogman

Date:

June 27, 2008 09:19

Subject:

Re: my $@ - a proposal

Message ID:

20080627161923.GB23537@woobling.org

On Fri, Jun 27, 2008 at 17:54:29 +0200, Steffen Mueller wrote:

> If one created a pragma to enable this behaviour lexically, one could
> use the same logic for loading it by default with use 5.012 as I
> proposed with my use-strict-by-default patch. Unfortunately, the
> implications with respect to what happens in -E weren't clear since
> those essentially do a "use 5.0CURRENT" but probably shouldn't add any
> new behaviour. Eventhough that change in behaviour was mostly liked, the
> patch hasn't made it in for this very reason. Rafael said he's still
> thinking about it.
> 
> For checking whether to create a lexical or not, you could then use a
> hint. So the plan would be:
> 
> - Make it possible to declare $@ as a lexical at all: "Can't use global
> $@ in "my" at -e line 1, near "my $@""
> - patch the core eval BLOCK implementation to create a lexical if the
> hint is set
> - create a simple pragma to set/unset the hint
> - apply aforementioned strict-by-default patch with
> s/strict/lexicalExceptions/g
> 
> One potential gotcha might be that if using a hints *hash* entry, the
> hash lookup on each eval {} could be considered large. I'm not sure, though.

This can be at compile time, such that having a lexical $@ in scope
compiles a variant of the eval {} opcode instead of the normal one.

In fact, I think I can maybe hack a PL_check based hack that makes
this work with any lexical.

This would be a POC in terms of API, due to the can't use global in
my problem, but should give us everything else:

	use lexical_error '$e'; # name of error var

	my $e; # if this is in scope it's used instead

the PL_check hack would go trrough all eval (string and block) and
replace them with wrapped versions that temporarily swap the error
SV with the one of the lexical one, and make sure that the jump exit
changes this back to what it was at the right time. This way 'die'
writes to the some SV, but that SV might not necessarily be $@.

If $@ is a normal var and not magical then this shouldn't be too
difficult and maybe a real patch can be made out of it.

-- 
  Yuval Kogman <nothingmuch@woobling.org>
https://nothingmuch.woobling.org  0xEBD27418

Thread Previous | Thread Next