From @moritz

09​:45 <@​moritz_> rakudo​: sub m (&f) { say "a" ~~ m/<f>/ }; regex outer {
a };
  m(&a)
09​:45 < p6eval> rakudo 7f8ba6​: OUTPUT«Null PMC access in isa()␤current
instr.​:
  '!SIGNATURE_BIND' pc 2559 (src/classes/Signature.pir​:377)␤»

note that there's no rule named 'a' defined.

A more helpful error message would be appropriate.

Moritz

--
Moritz Lenz
http://perlgeek.de/ | http://perl-6.de/ | http://sudokugarden.de/

From @bacek

On Wed Feb 25 00​:49​:42 2009, mlenz@​physik.uni-wuerzburg.de wrote​:

09​:45 <@​moritz_> rakudo​: sub m (&f) { say "a" ~~ m/<f>/ }; regex outer {
a };
m(&a)
09​:45 < p6eval> rakudo 7f8ba6​: OUTPUT«Null PMC access in isa()␤current
instr.​:
'!SIGNATURE_BIND' pc 2559
(src/classes/Signature.pir​:377)␤»

note that there's no rule named 'a' defined.

A more helpful error message would be appropriate.

Moritz

Patch available at
http://github.com/bacek/rakudo/commit/6a68c91cf0311df9472886caefef750fb2503df4

Now it produces this​:
$ ./perl6 -e 'sub m (&f) { say "a" ~~ m/<f>/ }; regex outer { a }; m(&a)'
Non-Callable argument for f in call to m

--
Bacek

The RT System itself - Status changed from 'new' to 'open'

From @pmichaud

On Fri, Feb 27, 2009 at 02​:21​:59AM -0800, Vasily Chekalkin via RT wrote​:

On Wed Feb 25 00​:49​:42 2009, mlenz@​physik.uni-wuerzburg.de wrote​:

09​:45 <@​moritz_> rakudo​: sub m (&f) { say "a" ~~ m/<f>/ }; regex outer {
a };
m(&a)
09​:45 < p6eval> rakudo 7f8ba6​: OUTPUT«Null PMC access in isa()␤current
instr.​:
'!SIGNATURE_BIND' pc 2559
(src/classes/Signature.pir​:377)␤»

note that there's no rule named 'a' defined.

A more helpful error message would be appropriate.

Patch available at
http://github.com/bacek/rakudo/commit/6a68c91cf0311df9472886caefef750fb2503df4

The patch conflates the concepts of "null" and "defined".
Instead of checking for a defined value, it should be checking
for a null value and doing the right thing after that.

This is also related to RT #​62948, where null values appear
in arrays and other aggregates. Instead of adding a bunch
of "is this value null" checks throughout the code, I'd
prefer to look for a way to automatically promote nulls
into undefs at the appropriate points.

Pm

From @pmichaud

On Tue, Mar 03, 2009 at 10​:21​:33AM +1100, Vasily Chekalkin wrote​:

Patrick R. Michaud wrote​:

This is also related to RT #​62948, where null values appear in arrays and other aggregates. Instead of adding a bunch of "is this value null" checks throughout the code, I'd prefer to look for a way to automatically promote nulls into undefs at the appropriate points.

Is it feasible to catch all "Null PMC access" exceptions at the top-level,
promote PMCNULL to Undef (or Failure) and resume from exception?

Good thinking... but there's not anything to "promote" -- i.e.,
from an exception handler we wouldn't have the ability to change
the register that has the PMCNULL.

Pm

From @pmichaud

On Tue, Mar 03, 2009 at 11​:01​:22AM +1100, Vasily Chekalkin wrote​:

Patrick R. Michaud wrote​:

Is it feasible to catch all "Null PMC access" exceptions at the
top-level, promote PMCNULL to Undef (or Failure) and resume from
exception?

Good thinking... but there's not anything to "promote" -- i.e., from an
exception handler we wouldn't have the ability to change the register that
has the PMCNULL.

Ouch. Looks like '!NULLORUNDEF' subroutine which returns boolean one of
few possible variants?

No, we can clean it up better at the source than we can by spreading
lots of NULLORUNDEF calls throughout the codebase.

Pm

From @bacek

Patrick R. Michaud wrote​:

Is it feasible to catch all "Null PMC access" exceptions at the
top-level, promote PMCNULL to Undef (or Failure) and resume from
exception?

Good thinking... but there's not anything to "promote" -- i.e., from an
exception handler we wouldn't have the ability to change the register that
has the PMCNULL.

Ouch. Looks like '!NULLORUNDEF' subroutine which returns boolean one of few
possible variants?

--
Bacek

From @bacek

Patrick R. Michaud wrote​:

This is also related to RT #​62948, where null values appear in arrays and other aggregates. Instead of adding a bunch of "is this value null" checks throughout the code, I'd prefer to look for a way to automatically promote nulls into undefs at the appropriate points.

Is it feasible to catch all "Null PMC access" exceptions at the top-level, promote PMCNULL to Undef (or Failure) and resume from exception?

--
Bacek

@moritz - Status changed from 'open' to 'resolved'