VK>> Test on Win32 pass.
VK>> Test on Linux will do within couple of days.
VK>> I cannot do tests on VMS, MacOS, OS2, but there should be nothing broken.

VK>> (it is still possible to temporarily lie about $^O and at least check
VK>> whether there will be syntax errors in generated Dynaloader.pm)

While performing some investigations, I discovered that cyrrently
DynaLoader_pm.PL contains some logic about environment variable
PERL_BUILD_EXPAND_CONFIG_VARS but this logic is currently broken.

A Dynaloader.pm script generated while
$ENV{PERL_BUILD_EXPAND_CONFIG_VARS} is true doomed to fail (even
fithout my patch!)

It seems to me that no-one uses this env var, that is why it is
stale and it is better be just removed.

I will do removing of PERL_BUILD_EXPAND_CONFIG_VARS with next patch to
Dynaloader (yet lib_pm.PL contains it) in case no-one objects.

If interesting, try running building and test process with that
environment variable set!
There is a mess with $pthsep variable, but not only that variable.

A script for playing with different $^O is simple:

for my $O (qw(MacOS MSWin32 darwin os2 VMS cygwin linux sunos)) {
  delete $ENV{PERL_BUILD_EXPAND_CONFIG_VARS};
  system("../../perl", '-we', "\$^O='$O';do './DynaLoader_pm.PL'");
  `mv DynaLoader.pm test0-DynaLoader-$O.pm`;
  $ENV{PERL_BUILD_EXPAND_CONFIG_VARS} = 1;
  system("../../perl", '-we', "\$^O='$O';do './DynaLoader_pm.PL'");
  `mv DynaLoader.pm test1-DynaLoader-$O.pm`;
}

Please disregard my previous dynaloader patch and instead use current
one, it should be better WRT $ENV{PERL_BUILD_EXPAND_CONFIG_VARS}.


VK>> Additionally, I dare to ask community to allow me to remove AUTOLOAD
VK>> mechanic for Dynaloader, because IMHO it makes things a bit heavier
VK>> and a bit more complex without good gain.

if no-one will argue against this, I will send a patch for this.


Best regards,
Vadim.

Attachment: DynaLoader_pm.PL.diff
Description: Binary data

Reply via email to