use and require are affected by the open pragma #11541
Comments
From @LeontThis is a bug report for perl from fawaka@gmail.com, use and require are affected by the open pragma, as seen here: % perl -e 'print "package F;\n # \xF1\n;1;"' > x.pl Flags: Site configuration information for perl 5.14.1: Configured by leon at Tue Jun 28 15:59:52 CEST 2011. Summary of my perl5 (revision 5 version 14 subversion 1) configuration: Platform: Locally applied patches: @INC for perl 5.14.1: Environment for perl 5.14.1: |
From @iabynOn Sun, Jul 31, 2011 at 02:05:23PM -0700, Leon Timmermans wrote:
Well, the pragma's docs says that it affects "all I/O" within the lexical -- |
|
The RT System itself - Status changed from 'new' to 'open' |
From @LeontOn Mon Aug 01 02:27:45 2011, davem wrote:
I really don't think loading a module is can be reasonably called «I/O Leon |
From @nwc10On Mon, Aug 01, 2011 at 10:27:12AM +0100, Dave Mitchell wrote:
$ cat >y.pl No warning. perlfunc.pod says: Otherwise, C<require> demands that a library file be included if it and makes no comment about lexical pragmata. The y.pl example above shows So it looks like the current behaviour is "as if" [and probably actually All things above being equal, I'm not sure if I think that this is a bug. require only loads a file once. Hence if you have two modules which require use open ":encoding(utf8)" and the other is no open; then what you actually get in %foo:: is going to depend on the order that Which is action at a distance. And feels horribly like it will cause much And caustic blog posts are bad marketing, even if we are so callous as to Or at least, that's how somewhat sleep-deprived me sees it. Nicholas Clark |
From @LeontOn Mon Aug 01 07:36:28 2011, nicholas wrote:
That's exactly what happens (see pp_require and check_type_and_open in
Yeah, that is another complication of this. The open pragma seems to be Leon |
From @HugmeirOn Mon, Aug 1, 2011 at 11:36 AM, Nicholas Clark <nick@ccl4.org> wrote:
Well, assuming this is considered a bug, here's a patch to resolve it : ) It Inline Patchdiff --git a/lib/open.t b/lib/open.t
index 1bf7dca..1d61086 100644
--- a/lib/open.t
+++ b/lib/open.t
@@ -186,18 +186,39 @@ SKIP: {
eval q[use Encode::Alias;use open ":std", ":locale"];
is($@, '', 'can use :std and :locale');
- use open IN => ':non-existent';
- eval {
- require Symbol; # Anything that exists but we havn't loaded
- };
- like($@, qr/Can't locate Symbol|Recursive call/i,
- "test for an endless loop in PerlIO_find_layer");
+#This test was using dubious behaviour - It relied on the open pragma to
END { # the test cases beyond __DATA__ need to be executed separately Inline Patchdiff --git a/pp_ctl.c b/pp_ctl.c
index d5924d8..7d5e58a 100644
--- a/pp_ctl.c
+++ b/pp_ctl.c
@@ -3649,7 +3649,7 @@ S_check_type_and_open(pTHX_ SV *name)
}
#if !defined(PERLIO_IS_STDIO) && !defined(USE_SFIO)
- return PerlIO_openn(aTHX_ NULL, PERL_SCRIPT_MODE, -1, 0, 0, NULL, 1,
|
From @Hugmeiropen_noprop.diffdiff --git a/lib/open.t b/lib/open.t
index 1bf7dca..1d61086 100644
--- a/lib/open.t
+++ b/lib/open.t
@@ -186,18 +186,39 @@ SKIP: {
eval q[use Encode::Alias;use open ":std", ":locale"];
is($@, '', 'can use :std and :locale');
- use open IN => ':non-existent';
- eval {
- require Symbol; # Anything that exists but we havn't loaded
- };
- like($@, qr/Can't locate Symbol|Recursive call/i,
- "test for an endless loop in PerlIO_find_layer");
+#This test was using dubious behaviour - It relied on the open pragma to propagate
+#it's effect into the layer used for require, which is no longer true.
+#See the next test as for why this was changed.
+
+# use open IN => ':non-existent';
+# eval {
+# require Symbol; # Anything that exists but we havn't loaded
+# };
+# like($@, qr/Can't locate Symbol|Recursive call/i,
+# "test for an endless loop in PerlIO_find_layer");
+
+
+ open my $out_fh, ">:raw", "x.pl";
+ print { $out_fh } "package F;\n # \xF1\n;1;";
+ close $out_fh;
+
+#This require would've thrown "utf8 "\xF1" does not map to Unicode" if require
+#was affected by open.
+ {
+ use open qw( :encoding(utf8) );
+ eval {
+ local @INC = ".";
+ require "x.pl";
+ };
+ is $@, '', "open does not propagate into require()";
+ }
}
END {
1 while unlink "utf8";
1 while unlink "a";
1 while unlink "b";
+ 1 while unlink "x.pl";
}
# the test cases beyond __DATA__ need to be executed separately
diff --git a/pp_ctl.c b/pp_ctl.c
index d5924d8..7d5e58a 100644
--- a/pp_ctl.c
+++ b/pp_ctl.c
@@ -3649,7 +3649,7 @@ S_check_type_and_open(pTHX_ SV *name)
}
#if !defined(PERLIO_IS_STDIO) && !defined(USE_SFIO)
- return PerlIO_openn(aTHX_ NULL, PERL_SCRIPT_MODE, -1, 0, 0, NULL, 1, &name);
+ return PerlIO_openn(aTHX_ ":raw", PERL_SCRIPT_MODE, -1, 0, 0, NULL, 1, &name);
#else
return PerlIO_open(p, PERL_SCRIPT_MODE);
#endif
|
From tchrist@perl.comJust because you call a function that does I/O doesn't mean do provides a separate lexical scope. A different question is $foo = `cat $file`. --tom |
From @LeontOn Mon, Aug 1, 2011 at 9:29 PM, Brian Fraser <fraserbn@gmail.com> wrote:
Perl modules/scripts are text files, not binary files. I don't think Leon |
From @nwc10On Mon, Aug 01, 2011 at 04:29:40PM -0300, Brian Fraser wrote:
Except I *think* that the call:
now opens the file then pushes "raw", and pushing "raw" does binmode() on the I'm not sure how to call PerlIO in such a way as to override the open pragma, PerlIO is not my favourite part of the core Perl source code, and hasn't been Nicholas Clark |
From @LeontOn Tue Aug 02 02:55:21 2011, nicholas wrote:
I think we should add a :default pseudo-layer. It should be fairly
It's nobody's favorite part of the core, which is its main problem. Leon |
From @cpansproutOn Tue Aug 02 02:55:21 2011, nicholas wrote:
How about temporarily clearing the hints (after using a SAVE* macro)?
|
From @cpansproutOn Tue Aug 02 06:21:58 2011, sprout wrote:
Or pushing the new scope before opening the file?
|
From @janduboisOn Tue, 02 Aug 2011, Leon Timmermans wrote:
No, it won't. It will switch the DATA filehandle to :raw mode as well, Perl actually used to read the source in :raw mode on Windows, but https://rt-archive.perl.org/perl5/Public/Bug/Display.html?id=28106 For Perl 5.14 I changed the default to reading source code in text mode, Cheers, |
From @HugmeirOn Tue, Aug 2, 2011 at 3:36 PM, Jan Dubois <jand@activestate.com> wrote:
Alright, then I think that Father C's suggestion is our winner here. Unfortunately I can't get SAVEHINTS(); to play nice (my ignorance of PerlIO |
From @LeontOn Tue, Aug 2, 2011 at 2:47 PM, Leon Timmermans via RT
Attached the patch. It turned out to be quite simple. Leon |
From @Leontdefault-perlio.patchdiff --git a/perlio.c b/perlio.c
index 72a71fe..c8d7141 100644
--- a/perlio.c
+++ b/perlio.c
@@ -837,6 +837,9 @@ PerlIO_get_layers(pTHX_ PerlIO *f)
* XS Interface for perl code
*/
+#define PerlIO_default_top() PerlIO_default_layer(aTHX_ -1)
+#define PerlIO_default_btm() PerlIO_default_layer(aTHX_ 0)
+
PerlIO_funcs *
PerlIO_find_layer(pTHX_ const char *name, STRLEN len, int load)
{
@@ -844,6 +847,8 @@ PerlIO_find_layer(pTHX_ const char *name, STRLEN len, int load)
IV i;
if ((SSize_t) len <= 0)
len = strlen(name);
+ if (len == 7 && strEQ(name, "default"))
+ return PerlIO_default_top();
for (i = 0; i < PL_known_layers->cur; i++) {
PerlIO_funcs * const f = PL_known_layers->array[i].funcs;
if (memEQ(f->name, name, len) && f->name[len] == 0) {
@@ -1246,9 +1251,6 @@ PerlIO_default_layer(pTHX_ I32 n)
return PerlIO_layer_fetch(aTHX_ av, n, PERLIO_FUNCS_CAST(&PerlIO_stdio));
}
-#define PerlIO_default_top() PerlIO_default_layer(aTHX_ -1)
-#define PerlIO_default_btm() PerlIO_default_layer(aTHX_ 0)
-
void
PerlIO_stdstreams(pTHX)
{
|
From @HugmeirOn Sun, Aug 7, 2011 at 6:42 PM, Leon Timmermans <fawaka@gmail.com> wrote:
With 5.15.2 out of the door, can this get committed? |
From @cpansproutOn Sat Aug 20 21:37:43 2011, fraserbn@gmail.com wrote:
If this adds a new pseudo-layer accessible from Perl space, then it |
From @HugmeirOn Sun, Aug 21, 2011 at 3:37 AM, Father Chrysostomos via RT <
Would something like this do? Inline Patchdiff --git a/lib/PerlIO.pm b/lib/PerlIO.pm
index 46e6e44..2c36d72 100644
--- a/lib/PerlIO.pm
+++ b/lib/PerlIO.pm
@@ -190,6 +190,15 @@ On Win32 platforms this I<experimental> layer uses the
+=item :default =head2 Custom Layers |
From @LeontOn Sun, Aug 21, 2011 at 8:37 AM, Father Chrysostomos via RT
At least as important, it needs tests. I wrote some, but I bumped into I think it'd be better to fix that first before applying this. Leon |
From @cpansproutOn Thu Aug 04 01:11:45 2011, fraserbn@gmail.com wrote:
Except it can’t work properly, because SAVEHINTS and PL_hints deal with PL_curcop->cop_hints can’t be modified, because that won’t be thread-safe.
|
From @HugmeirOn Mon, Aug 22, 2011 at 6:34 PM, Father Chrysostomos via RT <
Bit of a surprising discovery; Turns out that passing a single ":" for the That will only get the relevant default layers, while removing the colons Inline Patchdiff --git a/pp_ctl.c b/pp_ctl.c
index 3a2de27..a9c0afd 100644
--- a/pp_ctl.c
+++ b/pp_ctl.c
@@ -3692,7 +3692,7 @@ S_check_type_and_open(pTHX_ SV *name)
#if !defined(PERLIO_IS_STDIO) && !defined(USE_SFIO) (tests in a previous mail) Is this worth documenting somewhere? |
From @rgarciaOn 22 September 2011 16:29, Brian Fraser <fraserbn@gmail.com> wrote:
I vote yes, because I find this accidental feature actually neat. And |
From tchrist@perl.comRafael Garcia-Suarez <rgs@consttype.org> wrote
My instinct says that you're right. It is always cool to find new uses I *REALLY* wish I understood layers better. I can do simple stuff, thanks, --tom |
From @rgarciaOn 22 September 2011 17:09, Tom Christiansen <tchrist@perl.com> wrote:
There's the scary low-level docs in perliol. |
From @cpansproutOn Thu Sep 22 07:30:00 2011, fraserbn@gmail.com wrote:
Thank you. Applied as 639dfab.
I didn’t read that sentence until I had already written one. It turns |
From @cpansproutOn Thu Sep 22 07:30:00 2011, fraserbn@gmail.com wrote:
Yes, and I have documented it with commit c0fd9d2. |
|
@cpansprout - Status changed from 'open' to 'resolved' |
From @LeontOn Thu, Sep 22, 2011 at 5:09 PM, Tom Christiansen <tchrist@perl.com> wrote:
Barely. perliol defines some internals but that's not useful for Most of it is reading the source and trying things out. That's also Leon |
Migrated from rt.perl.org#96008 (status was 'resolved')
Searchable as RT96008$
The text was updated successfully, but these errors were encountered: