Not OK: perl 5.00561 on PA-RISC1.1 10.20

[p5pRT]

Migrated from rt.perl.org#1527 (status was 'resolved')

Searchable as RT1527$

[p5pRT]

From lewart@www.cvm.uiuc.edu

---

Perl5-Porters,

getpwent and getpwuid are broken, but getpwnam is okay​:
1) getpwent returns the last entry in /etc/passwd .
2) getpwuid($num) returns the last entry in /etc/passwd if $num is
  a valid UID; otherwise it returns an empty string.

Why didn't "make test" catch such egregious errors? Because of the
following stanza from t/op/pwent.t​:
  94​: # Protect against one-to-many and many-to-one mappings.
  95​: if ($name_s ne $name) {
  96​: @​n = getpwnam($name_s);
  97​: ($name,$passwd,$uid,$gid,$quota,$comment,$gcos,$home,$shell) = @​n;
  98​: next if $name_s ne $name;
  99​: }
What is the point of the above clause?

Cheers,
Daniel Lewart
d-lewart@​uiuc.edu

---

---

Site configuration information for perl 5.00561​:

Configured by lewart at Mon Sep 6 14​:01​:38 CDT 1999.

Summary of my perl5 (revision 5.0 version 5 subversion 61) configuration​:
  Platform​:
  osname=hpux, osvers=10.20, archname=PA-RISC1.1
  uname='hp-ux www b.10.20 a 9000715 2013547832 two-user license '
  config_args='-de -Dcc=gcc -Dprefix=/ext/perl5.005_61'
  hint=recommended, useposix=true, d_sigaction=define
  usethreads=undef useperlio=undef d_sfio=undef
  use64bits=undef usemultiplicity=undef
  Compiler​:
  cc='gcc', optimize='-O', gccversion=2.95.1 19990816 (release)
  cppflags='-D_HPUX_SOURCE -I/usr/local/include'
  ccflags ='-D_HPUX_SOURCE -I/usr/local/include'
  stdchar='unsigned char', d_stdstdio=define, usevfork=false
  intsize=4, longsize=4, ptrsize=4, doublesize=8
  d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=8
  alignbytes=8, usemymalloc=y, prototype=define
  Linker and Libraries​:
  ld='ld', ldflags =' -L/usr/local/lib'
  libpth=/usr/local/lib /lib/pa1.1 /lib /usr/lib /usr/ccs/lib
  libs=-lnsl_s -lndbm -lgdbm -ldb -ldld -lm -lc -lndir -lcrypt -lsec
  libc=/lib/libc.sl, so=sl, useshrplib=false, libperl=libperl.a
  Dynamic Linking​:
  dlsrc=dl_hpux.xs, dlext=sl, d_dlsymun=undef, ccdlflags='-Wl,-E -Wl,-B,deferred '
  cccdlflags='-fpic', lddlflags='-b -L/usr/local/lib'

Locally applied patches​:
 

---

@​INC for perl 5.00561​:
  /ext/perl5.005_61/lib/5.00561/PA-RISC1.1
  /ext/perl5.005_61/lib/5.00561
  /ext/perl5.005_61/lib/site_perl/5.00561/PA-RISC1.1
  /ext/perl5.005_61/lib/site_perl
  .

---

Environment for perl 5.00561​:
  HOME=/home/lewart
  LANG (unset)
  LANGUAGE (unset)
  LD_LIBRARY_PATH (unset)
  LOGDIR (unset)
  PATH=/usr/local/bin​:/usr/bin​:/usr/ccs/bin​:/usr/contrib/bin​:/opt/nettladm/bin​:/opt/pd/bin​:/usr/bin/X11​:/usr/contrib/bin/X11​:/opt/upgrade/bin
  PERL_BADLANG (unset)
  SHELL=/usr/bin/ksh
  SHLIB_PATH (unset)

[p5pRT]

From @jhi

Daniel S. Lewart writes​:

getpwent and getpwuid are broken, but getpwnam is okay​:
1) getpwent returns the last entry in /etc/passwd .
2) getpwuid($num) returns the last entry in /etc/passwd if $num is
a valid UID; otherwise it returns an empty string.

Huh? This sounds so weird that I have to check​: getpwent() *always*
returns only the last entry in /etc/passwd()? After
endpwent()+setpwent(), too?

Ditto for getpwuid($valid_num).

The "empty string" case of getpwuid($invalid_num) is as it should be.
If you look closer you'll notice it's the undef (or an empty list
in array context).

(I checked almost-_62 in Digital UNIX and getpwent() and getpwuid()
work okay. I'll try compiling the same Perl tomorrow--but I don't
have HP-UX 10.20 anymore, I have 11.0, but I don't have gcc in there,
only the HP ANSI cc.)

Why didn't "make test" catch such egregious errors? Because of the
following stanza from t/op/pwent.t​:
94​: # Protect against one-to-many and many-to-one mappings.
95​: if ($name_s ne $name) {
96​: @​n = getpwnam($name_s);
97​: ($name,$passwd,$uid,$gid,$quota,$comment,$gcos,$home,$shell) = @​n;
98​: next if $name_s ne $name;
99​: }
What is the point of the above clause?

The passwd file might have both one username -> many uids mappings
(illegal but possible) and one uid -> many usernames mappings (slightly
less illegal, but sometimes useful). We cannot do many (any?) useful
checks if we have any many-mappings.

I must be slow but could you elaborate on how the above test makes
getpw*() tests to fail (to give a false positive)?

--
$jhi++; # http​://www.iki.fi/jhi/
  # There is this special biologist word we use for 'stable'.
  # It is 'dead'. -- Jack Cohen

[p5pRT]

From @jhi

Daniel S. Lewart writes​:

This is a build failure report for perl from d-lewart@​uiuc.edu,
generated with the help of perlbug 1.27 running under perl 5.00561.

-----------------------------------------------------------------
Perl5-Porters,

getpwent and getpwuid are broken, but getpwnam is okay​:
1) getpwent returns the last entry in /etc/passwd .
2) getpwuid($num) returns the last entry in /etc/passwd if $num is
a valid UID; otherwise it returns an empty string.

Soon-to-be-62 in HP-UX 11.0 compiled with gcc 2.8.1​: getpwent() most
certainly doesn't return the last entry in /etc/passwd -- not until
the last one is supposed to be returned.
getpwuid($valid_num) works okay, too.

--
$jhi++; # http​://www.iki.fi/jhi/
  # There is this special biologist word we use for 'stable'.
  # It is 'dead'. -- Jack Cohen