Carp::croak doesn't quite replace die - missing <fh> line # phrase #11569
Comments
From root@overkill.sb.litts.netCreated by tlhackque@yahoo.comCarp::croak is documented in the camel book as "working like die does, except misdirecting blame to the caller." Perldoc says "croak - die of errors (from perspective of caller)". This is almost true, however careful inspection reveals that croak does not include file position information ( ", <fh> line $." ) as die does (when it can). It turns out that it *is* possible to determine the current file handle (although I wish it were easier). Carp should do this and add it to the traceback, as it really is useful information. Here's how to generate the missing phrase: $msg = join( '', @_ ) . " at $script line $n"; (add traceback) For extra credit, it would be nice to have an better mechanism to get the current file handle name bound to Test cases (type <cr> to get the output): a) die's behavior: xx at -e line 1, <STDIN> line 1. b) Carp::croak's behavior: xx at -e line 1 Thanks for your consideration. Perl Info |
From @jkeenanOn Tue Aug 09 15:53:13 2011, root@overkill.sb.litts.net wrote:
Would it be possible for you to prepare this as a patch to Thank you very much. |
|
The RT System itself - Status changed from 'new' to 'open' |
From @doyOn Wed, Dec 14, 2011 at 03:45:10PM -0800, James E Keenan via RT wrote:
It is also not quite equivalent to die because it is missing the $ perl -E'die foo' I wonder if fixing this would be worth it or not. -doy |
From @cpansproutOn Wed Dec 14 15:55:20 2011, doy@tozt.net wrote:
That has bothered me, too. I think that could be considered a bug fix, -- Father Chrysostomos |
From tlhackque@yahoo.comOn Wed Dec 14 15:45:10 2011, jkeenan wrote:
Sure, attached. Against v5.12.4. It's as in the original report, Note this also fixed the missing '.'. # perl -MCarp -e 'my $x=<STDIN>; sub foo { croak "xx";}; foo();' xx at -e line 1, <STDIN> line 1. Enjoy. |
From tlhackque@yahoo.comCarpFH.patch--- /usr/share/perl5/Carp.pm~ 2011-10-05 12:22:51.000000000 -0400
+++ /usr/share/perl5/Carp.pm 2011-12-14 20:53:52.082610049 -0500
@@ -199,11 +199,21 @@
my $tid = threads->tid;
$tid_msg = " thread $tid" if $tid;
}
my %i = caller_info($i);
- $mess = "$err at $i{file} line $i{line}$tid_msg\n";
+ $mess = "$err at $i{file} line $i{line}$tid_msg";
+ if( defined $. ) {
+ local $@ = '';
+ eval {
+ die;
+ };
+ if($@ =~ /^Died at .*(, <.*?> line \d+).$/ ) {
+ $mess .= $1;
+ }
+ }
+ $mess .= ".\n";
while (my %i = caller_info(++$i)) {
$mess .= "\t$i{sub_name} called at $i{file} line $i{line}$tid_msg\n";
}
|
From @davidnicol
Conversely, "die" retains an unnecessary period which should be removed for |
From @cpansproutOn Wed Dec 14 21:11:34 2011, davidnicol@gmail.com wrote:
I notice you are not following the cleaner style in your prose -- Father Chrysostomos |
From @davidnicolOn Wed, Dec 14, 2011 at 11:32 PM, Father Chrysostomos via RT <
Am I the only one who has ever imagined that dot to be there to separate |
From tlhackque@yahoo.comI reported that croak doesn't replace die as documented. die has the dot; the patch that I provided actually matches on that dot As for croak; I really don't care about this nit. I put the dot into Personally, I think the energy spent arguing over this could be better I'll leave resolution of this in your hands; I've done what I can, and |
From [Unknown Contact. See original ticket]I reported that croak doesn't replace die as documented. die has the dot; the patch that I provided actually matches on that dot As for croak; I really don't care about this nit. I put the dot into Personally, I think the energy spent arguing over this could be better I'll leave resolution of this in your hands; I've done what I can, and |
From @cpansproutOn Thu Dec 15 01:20:45 2011, tlhackque wrote:
Believe me, I’ve been thinking about that. $$. can’t work, as that
I’ve applied your patch without the dot in the interest of getting this Thank you. I still think we should add the dot, but shouldn’t it apply to every BTW, we have a policy that every contributor be listed in the AUTHORS -- Father Chrysostomos |
From tlhackque@yahoo.comOn Thu Dec 15 22:20:38 2011, sprout wrote:
I guess it's a question of how much magic to apply. The thought behind ${^LAST_FH} or some such would work as a name. For efficiency, perhaps Obviously, once one has the name, it can be symbolically de-referenced Documentation for such a variable should probably note that the last FH
I guess that would be more consistent, but slightly riskier. I felt
I'm OK either way. My real time is Timothe Litt, and my stable e-mail |
From @cpansproutOn Fri Dec 16 03:06:17 2011, tlhackque wrote:
Because there is a variable named $$, punctuation variables need braces $. is magical, but it works like a tied variable (as do most [all?] (I know I’m being verbose, but I thought you might want to see my
I like that name.
I don’t understand that last sentence. If ${^LAST_FH} is a typeglob (not a problem, BTW, since typeglobs are
Not necessarily. open my $fh... where $fh is undef creates a new
Certainly.
Speaking of controversy, the deadline for controversial changes before
Added with commit 5aaeb4d.
Considering your use of underscores, I did not change the e-mail address -- Father Chrysostomos |
From tlhackque@yahoo.comOn Fri Dec 16 08:39:20 2011, sprout wrote:
Fair enough. You're doing the work, your reasoning wins.
Sorry, I'll expand. Since (as noted later), the LAST_FH can't be The alternative would be to make @{^LAST_FH} return something like (
|
From tlhackque@yahoo.comSorry, my reply escaped before I was done. On Fri Dec 16 09:12:46 2011, tlhackque wrote:
Fair enough. You're doing the work, your reasoning wins.
Sorry, I'll expand. Since (as noted later), the LAST_FH can't be The alternative would be to make @{^LAST_FH} return something like (
As noted above, you only have $. if the LAST_FH is select()ed.
Returning the FH as the third element of a list would handle all these.
OK, that's in your hands. I wasn't aware of the schedule. Maybe it
I was more worried about e-mail harvesting on this (rt) site than the I appreciate all your effort on this. |
From @doyOn Fri, Dec 16, 2011 at 09:26:01AM -0800, tlhackque via RT wrote:
Alternatively, ${^LAST_FH}->input_line_number. -doy |
From tlhackque@yahoo.comOne more thought. Returning an array would seem to allow using @. as the variable name. The use english version could be @{^LAST_FH}. |
From [Unknown Contact. See original ticket]One more thought. Returning an array would seem to allow using @. as the variable name. The use english version could be @{^LAST_FH}. |
From @cpansproutOn Fri Dec 16 09:26:01 2011, tlhackque wrote:
I’ll admit I had to check, but $. is bound to the last-read handle,
I’m afraid I have to disagree with you there. Having a special array
OK. I’ve changed the e-mail address with commit 9fcef2a. -- Father Chrysostomos |
From @apSo I’m seeing a whole rigmarole for designing a special magical variable Then I see this: * tlhackque via RT <perlbug-followup@perl.org> [2011-12-16 18:45]:
And my first thought is: so why not simply make a `lastread` function It could be a weak keyword (preferably) or a feature (and for once Regards, |
From @cpansproutOn Fri Dec 16 18:13:18 2011, aristotle wrote:
Would it be an lvalue keyword, like pos? -- Father Chrysostomos |
From @davidnicolOn Fri, Dec 16, 2011 at 11:40 PM, Father Chrysostomos via RT <
What possible circumstance could require changing the identity of the On the other hand, if you could localize it, you could hide it and restore |
From @ap* Father Chrysostomos via RT <perlbug-followup@perl.org> [2011-12-17 06:45]:
Definite maybe. And it can always be made lvalue later if someone finds * David Nicol <davidnicol@gmail.com> [2011-12-17 08:20]:
It sits very oddly in `caller`’s return. Do you get different values depending on whether you ask for `caller 0` I don’t think that is the right place for it. Regards, |
From @cpansproutOn Sat Dec 17 03:16:09 2011, aristotle wrote:
I meant that almost as an ironic question. Having a special keyword Do we want to follow the interface of select(), and have Another, more troubling question: I have just discovered that I was -- Father Chrysostomos |
From @cpansproutAttached is a proof of concept. It still needs tests and docs. It will With it applied: $ ./perl -Ilib -le '<foo>; print ${^LAST_FH}; print \${^LAST_FH}; print So with a single new variable, you can get a real typeglob that you can Who could ask for anything more? Later we could even allowed it to be assigned to, such that ${^LAST_FH} -- Father Chrysostomos |
From @cpansproutInline Patchdiff --git a/gv.c b/gv.c
index 3a978f2..9bfea53 100644
--- a/gv.c
+++ b/gv.c
@@ -1818,6 +1818,10 @@ Perl_gv_fetchpvn_flags(pTHX_ const char *nambeg, STRLEN full_len, I32 flags,
if (strEQ(name2, "LOBAL_PHASE"))
goto ro_magicalize;
break;
+ case '\014': /* $^LAST_FH */
+ if (strEQ(name2, "AST_FH"))
+ goto ro_magicalize;
+ break;
case '\015': /* $^MATCH */
if (strEQ(name2, "ATCH"))
goto magicalize;
diff --git a/mg.c b/mg.c
index c55ca63..b1865b7 100644
--- a/mg.c
+++ b/mg.c
@@ -882,6 +882,15 @@ Perl_magic_get(pTHX_ SV *sv, MAGIC *mg)
case '\011': /* ^I */ /* NOT \t in EBCDIC */
sv_setpv(sv, PL_inplace); /* Will undefine sv if PL_inplace is NULL */
break;
+ case '\014': /* ^LAST_FH */
+ if (strEQ(remaining, "AST_FH"))
+ sv_setsv_nomg(
+ sv,
+ PL_last_in_gv && isGV_with_GP(PL_last_in_gv)
+ ? (SV *)PL_last_in_gv
+ : NULL
+ );
+ break;
case '\017': /* ^O & ^OPEN */
if (nextchar == '\0') {
sv_setpv(sv, PL_osname); |
From @rjbs* Father Chrysostomos via RT <perlbug-followup@perl.org> [2011-12-16T11:39:21]
(I agree on both counts.) -- |
From @jkeenanOn Sun Dec 18 00:20:57 2011, sprout wrote:
Are we going to follow up on the approach taken in this patch? Tests? Thank you very much. |
From @cpansproutOn Fri Sep 14 18:29:29 2012, jkeenan wrote:
I had nearly forgotten about this. It was in the back of my mind, but I found out after writing that patch that I was mistaken in thinking my $globref = \*STDIN; But that only works if *STDIN{IO} is defined. Otherwise it dies. But So that leads me to ask? What is the best value for ${^LAST_FH}? My In either case, we are not following existing precedent. select() *${^LAST_FH}; # glob itself, possibly a strict error That third line is not pretty. If we always use a reference: *${^LAST_FH}; # glob itself # If ${^LAST_FH} is undef: If we make ${^LAST_FH} itself a typeglob: ${^LAST_FH}; # typeglob, but a copy of the # If ${^LAST_FH} is undef: Despite the errors with undef, I think I now prefer a reference. Any thoughts? I can probably cook up a proper patch once I have finished fixing the Er, when am I going to get to this? :-) -- Father Chrysostomos |
From @rjbs* Father Chrysostomos via RT <perlbug-followup@perl.org> [2012-09-15T01:28:10]
Same. -- |
From @cpansproutOn Mon Sep 17 17:59:19 2012, perl.p5p@rjbs.manxome.org wrote:
I have now applied it as b932b26. Father Chrysostomos |
|
@cpansprout - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#96672 (status was 'resolved')
Searchable as RT96672$
The text was updated successfully, but these errors were encountered: