segfault (dead loop) with Encoding, use open :locale, print STDERR #8761
Comments
From aa29@mail.ruTo: perlbug@perl.org This is a bug report for perl from aa29@aa29xp, PerlIO: Warning with char that does not map to STDERR encoding leads binmode(STDERR, ":encoding(cp866)"); Then Perl tries to encode to cp866 while flushing STDERR buffer, it ext\Encode\Encode.xs, line 153: if (check & ENCODE_WARN_ON_ERR){ But "\302\253" already is in the STDERR buffer, and Perl_warner just Flags: Site configuration information for perl v5.8.4: Configured by aa29 at Mon May 17 17:59:46 2004. Summary of my perl5 (revision 5 version 8 subversion 4) configuration: Locally applied patches: @INC for perl v5.8.4: Environment for perl v5.8.4: PATH=C:\cygwin\bin;C:\Tcl\bin;C:\WINDOWS\system32;C:\WINDOWS;C:\WINDOWS\Syst aa29 |
From bulb@ucw.czCreated by bulb@ucw.czHello, I have recently seen perl crashing with segfault for certain combinations of #!/usr/bin/perl This script crashes on 3 different systems. One is SuSE 9.0 with perl 5.8.1, The script does NOT crash when STDERR is replaced by STDOUT, but rather "\x{8336}" does not map to iso-8859-2 at perlio-bug.pl line 9. as expected. Since the contents of the perlio layer stack might be important bulb@efreet:~$ perl -MData::Dumper -MPerlIO -e 'print Dumper(PerlIO::get_layers(STDERR, details => 1))' bulb@efreet:~$ perl -MData::Dumper -MPerlIO -e 'print Dumper(PerlIO::get_layers(STDOUT, details => 1))' bulb@efreet:~$ perl -MData::Dumper -MPerlIO -e 'binmode(STDERR, ":encoding(iso-8859-2)"); print Dumper(PerlIO::get_layers(STDERR, details => 1))' bulb@efreet:~$ perl -MData::Dumper -MPerlIO -e 'binmode(STDOUT, ":encoding(iso-8859-2)"); print Dumper(PerlIO::get_layers(STDOUT, details => 1))' This is a distribution perl, which is stripped, so the backtrace won't be (gdb) bt And this goes on for THOUSANDS of lines. It ends LOOOONG way down by: #66106 0x00002aaaab813a27 in XS_Encode__XS_encode () from /usr/lib/perl/5.8/auto/Encode/Encode.so Feel free to ask for more information. Regards, Jan Hudec <bulb@ucw.cz> Perl Info |
From bulb@ucw.czHello Again, I've played with the bug a little more and I include several more scripts, First, let's get rid of the dependence on UTF-8 in the mail. This is the #!/usr/bin/perl Now the next script does NOT crash. It works normally: #!/usr/bin/perl As you can see, the only difference is the line: $SIG{__WARN__} = sub { }; So it shows, that the problem is caused by printing to STDERR from a warning #!/usr/bin/perl Now if the print is not to STDERR, it works. The following script does not #!/usr/bin/perl The most surprising is this script though. The next script replaces STDERR #!/usr/bin/perl Now this script does crash, but in a DIFFERENT WAY. Instead of immediately "\x{8336}" does not map to ascii at perlio-bug4.pl line 3. And exits with status 2. Note, that this was on a terminal, where STDOUT is -- |
From @smpeters
The code is essentially going into an infinite loop. The code discovers Here's the fatal end of the stack trace. #0 0x080fc619 in Perl_sv_vcatpvfn (sv=0x933e5d8, |
|
The RT System itself - Status changed from 'new' to 'open' |
From pixel@mandriva.comThis is a bug report for perl from pixel@mandriva.com, this segfaults if the locale charset is not utf8: use open ':locale'; Flags: Site configuration information for perl v5.8.8: Configured by Mandriva at Fri Sep 8 20:00:54 CEST 2006. Summary of my perl5 (revision 5 version 8 subversion 8) configuration: Locally applied patches: @INC for perl v5.8.8: Environment for perl v5.8.8: |
From @eserte"pixel@mandriva.com (via RT)" <perlbug-followup@perl.org> writes:
This seems to be an endless loop. This part is repeated over and over #852 0x282f9fc8 in PerlIOEncode_flush (f=0x81c0014) at encoding.xs:419 Regards,
-- |
|
The RT System itself - Status changed from 'new' to 'open' |
From @eserte"pixel@mandriva.com (via RT)" <perlbug-followup@perl.org> writes:
Attached is a patch which provides a workaround for the problem by |
From @eserteperlio-encoding.patch#
#
# To apply this patch:
# STEP 1: Chdir to the source directory.
# STEP 2: Run the 'applypatch' program with this patch file as input.
#
# If you do not have 'applypatch', it is part of the 'makepatch' package
# that you can fetch from the Comprehensive Perl Archive Network:
# http://www.perl.com/CPAN/authors/Johan_Vromans/makepatch-x.y.tar.gz
# In the above URL, 'x' should be 2 or higher.
#
# To apply this patch without the use of 'applypatch':
# STEP 1: Chdir to the source directory.
# STEP 2: Run the 'patch' program with this file as input.
#
#### End of Preamble ####
#### Patch data follows ####
diff -up 'bleedperl/ext/PerlIO/encoding/encoding.xs' 'bleedperl3/ext/PerlIO/encoding/encoding.xs'
Index: ./ext/PerlIO/encoding/encoding.xs
Prereq: 0.3
--- ./ext/PerlIO/encoding/encoding.xs Tue Jun 13 21:29:10 2006
+++ ./ext/PerlIO/encoding/encoding.xs Mon Feb 5 22:41:59 2007
@@ -48,6 +48,7 @@ typedef struct {
SV *enc; /* the encoding object */
SV *chk; /* CHECK in Encode methods */
int flags; /* Flags currently just needs lines */
+ int inEncodeCall; /* trap recursive encode calls */
} PerlIOEncode;
#define NEEDS_LINES 1
@@ -147,6 +148,7 @@ PerlIOEncode_pushed(pTHX_ PerlIO * f, co
}
e->chk = newSVsv(get_sv("PerlIO::encoding::fallback", 0));
+ e->inEncodeCall = 0;
FREETMPS;
LEAVE;
@@ -404,6 +406,7 @@ PerlIOEncode_flush(pTHX_ PerlIO * f)
STRLEN len;
SSize_t count = 0;
if ((PerlIOBase(f)->flags & PERLIO_F_WRBUF) && (e->base.ptr > e->base.buf)) {
+ if (e->inEncodeCall) return 0;
/* Write case - encode the buffer and write() to layer below */
PUSHSTACKi(PERLSI_MAGIC);
SPAGAIN;
@@ -416,9 +419,12 @@ PerlIOEncode_flush(pTHX_ PerlIO * f)
XPUSHs(e->bufsv);
XPUSHs(e->chk);
PUTBACK;
+ e->inEncodeCall = 1;
if (call_method("encode", G_SCALAR) != 1) {
+ e->inEncodeCall = 0;
Perl_die(aTHX_ "panic: encode did not return a value");
}
+ e->inEncodeCall = 0;
SPAGAIN;
str = POPs;
PUTBACK;
@@ -453,6 +459,7 @@ PerlIOEncode_flush(pTHX_ PerlIO * f)
}
/* See if there is anything left in the buffer */
if (e->base.ptr < e->base.end) {
+ if (e->inEncodeCall) return 0;
/* Bother - have unread data.
re-encode and unread() to layer below
*/
@@ -472,9 +479,12 @@ PerlIOEncode_flush(pTHX_ PerlIO * f)
XPUSHs(str);
XPUSHs(e->chk);
PUTBACK;
+ e->inEncodeCall = 1;
if (call_method("encode", G_SCALAR) != 1) {
- Perl_die(aTHX_ "panic: encode did not return a value");
+ e->inEncodeCall = 0;
+ Perl_die(aTHX_ "panic: encode did not return a value");
}
+ e->inEncodeCall = 0;
SPAGAIN;
str = POPs;
PUTBACK;
#### End of Patch data ####
#### ApplyPatch data follows ####
# Data version : 1.0
# Date generated : Mon Feb 5 22:58:23 2007
# Generated by : makepatch 2.03
# Recurse directories : Yes
# Excluded files : (\A|/).*\~\Z
# (\A|/).*\.a\Z
# (\A|/).*\.bak\Z
# (\A|/).*\.BAK\Z
# (\A|/).*\.elc\Z
# (\A|/).*\.exe\Z
# (\A|/).*\.gz\Z
# (\A|/).*\.ln\Z
# (\A|/).*\.o\Z
# (\A|/).*\.obj\Z
# (\A|/).*\.olb\Z
# (\A|/).*\.old\Z
# (\A|/).*\.orig\Z
# (\A|/).*\.rej\Z
# (\A|/).*\.so\Z
# (\A|/).*\.Z\Z
# (\A|/)\.del\-.*\Z
# (\A|/)\.make\.state\Z
# (\A|/)\.nse_depinfo\Z
# (\A|/)core\Z
# (\A|/)tags\Z
# (\A|/)TAGS\Z
# v 'patchlevel.h' 4732 1160388693 33056
# p 'ext/PerlIO/encoding/encoding.xs' 16562 1170711719 0100440
#### End of ApplyPatch data ####
#### End of Patch kit [created: Mon Feb 5 22:58:23 2007] ####
#### Patch checksum: 99 3441 30793 ####
#### Checksum: 117 4065 16793 ####
|
From @eserte-- tkrevdiff - graphical display of diffs between revisions (RCS, CVS or SVN) |
From @rgarciaOn 05 Feb 2007 23:04:07 +0100, Slaven Rezic <slaven@rezic.de> wrote:
The patch might be interesting to prevent similar segfaults in the |
From @eserte"Rafael Garcia-Suarez" <rgarciasuarez@gmail.com> writes:
Yes, this is the problem, and this is what the patch tries to fix. The With the inEncodeCall flag I tried to prevent calling encode() again I am not sure if the patch is right, because Regards, -- tksm - Perl/Tk program for searching and replacing in multiple files |
From @rgarciaOn 06 Feb 2007 21:59:53 +0100, Slaven Rezic <slaven@rezic.de> wrote:
Thanks, applied as change #30213, adding also Pixel's test case. (and
I don't see why it wouldn't be.
I don't think so, since that wasn't designed for that apparently. |
|
@rgs - Status changed from 'open' to 'resolved' |
From @dcollinsnThis was fixed somewhere between 5.8.8 and 5.8.9 by this monster of a commit: f266b49 is the first bad commit Integrate: [ 28766] [ 28798] [ 28799] [ 28903] [ 29173] [ 29702] [ 29751] [ 30213] [ 30214] [ 30233] p4raw-id: //depot/maint-5.8/perl@30340 :100644 100644 fb9a35503bbd64f925752279f2d4d92ae4aec465 4b59564147fb0caec127a3dcdccdca435cf74027 M MANIFEST Test to follow. -- |
From @dcollinsnThis (RT #37234) is, in fact, a duplicate of RT #41442. 41442 /has/ a test, and their test is better than my test. So, merging and closing. -- |
|
@dcollinsn - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#41442 (status was 'resolved')
Searchable as RT41442$
The text was updated successfully, but these errors were encountered: