Front page | perl.perl5.porters |
Postings from November 1999
Hi!
> > No, the scenario is as follows:
> >
> > 1) perl_parse() saves the pointer to the environment in PL_origenviron
> > 2) some extension interface (here Term::ReadLine::Gnu) is called
> > and does modify the pointer to the environment via putenv(3)
> > 3) the perlscript tries to assign a new value to an environment
> > variable and perl therefore calls my_setenv()
> > 4) my_setenv() does check if it needs to copy the environment but
> > the check returns false since (environ != PL_origenviron)
> > 5) my_setenv() calls Safefree() and free()s some unallocated memory
>
> *What* it calls Safefree() on?
Well, the appended perl-script-testcase had four lines of code, one of them is
$ENV{PATH} ="foo";
so the Safefree() has been called on environ[i] with i pointing PATH.
The testcase dumps only core if LINES and COLUMNS aren't defined in the
env before starting the script so that the putenv() called by
set_lines_and_columns() in readline-4.0:shell.c has to call malloc() and
therefore modifies the global var char **environ.
> How can it be that this is unallocated?
Since it has never been malloc()ed by perl due to the weak test in
my_setenv()?
> Judging by what you wrote, you are running some old version of Perl,
> such as 5.005_03.
This is the stable version, can't be that old?
> In 5.005_50 I can see that the only Safefree() is
> done on environ[i]. Due to documentation of putenv() I can see,
> environ[i] *should be* malloced.
But the free() is called with the pointer to a memory block which has
never been malloc()'ed by perl.
> I think you need to concentrate on these questions.
free()ing some memory which has never been malloc()ed is asking
for trouble. IMHO this is out of question.
Kind regards,
Joerg
--
Gaertner Datensysteme 38114 Braunschweig
Joerg Schumacher Hamburger Str. 273a
Tel: 0531-2335555 Fax: 0531-2335556