Re: testing blead on seeds of Mac OS X 10.5 Leopard

On 20/10/2007, Darren Duncan <darrendarrenduncan.net> wrote:

> Doing that spat out reams of text, with the last line
saying
>
>    None of your locales were broken.
>
> But a few lines prior to that, it said:
>
>    The locale of definition
>
>      be_BY.CP1131
>
>    on your system may have errors because the locale
test 99 failed on
> that locale.

OK, so that's the same bug.

> >It would be also helpful to have a verbose run of
op/pwent.t and see
> >why it's failing.
>
> Running this:
>
>    cd t && ./perl -TI . -MTestInit
../op/pwent.t
>
> Resulted in this immediate death:
>
>    Insecure $ENV while running with -T switch at
op/pwent.t line 22.
>    BEGIN failed--compilation aborted at op/pwent.t line
68.

Well, take out the -TI. : I added it to the command-line for
locale.t,
because this test has a -T on the #!perl -line. But it's not
needed
for pwent.t, and actually can be harmful, as you saw.

At 9:40 AM +0200 10/21/07, Rafael Garcia-Suarez wrote:
>On 20/10/2007, Darren Duncan <darrendarrenduncan.net> wrote:
>  > Running this:
>>
>  >    cd t && ./perl -TI . -MTestInit
../op/pwent.t
>>
>>  Resulted in this immediate death:
>>
>>     Insecure $ENV while running with -T
switch at op/pwent.t line 22.
>>     BEGIN failed--compilation aborted at op/pwent.t
line 68.
>
>Well, take out the -TI. : I added it to the command-line
for locale.t,
>because this test has a -T on the #!perl -line. But it's
not needed
>for pwent.t, and actually can be harmful, as you saw.

Okay, so I did another try, this time with the same blead I
used last time.

Running this:

     cd t && ./perl -I . -MTestInit ../op/pwent.t

Resulted in this:

     1..2
     # where /etc/passwd
     # max = 25, n = 25, perfect = 0
     #
     # The failure of op/pwent test is not necessarily
serious.
     # It may fail due to local password administration
conventions.
     # If you are for example using both NIS and local
passwords,
     # test failure is possible.  Any distributed password
scheme
     # can cause such failures.
     #
     # What the pwent test is doing is that it compares the
26 first
     # entries of /etc/passwd
     # with the results of getpwuid() and getpwnam() call. 
If it finds no
     # matches at all, it suspects something is wrong.
     #
     not ok 1	# (not necessarily serious: run t/op/pwent.t
by itself)
     ok 2

Running this:

     ./perl t/op/pwent.t

Resulted in the same thing.

So this appears to be non-serious, according to the output.

This all said, I do recall from reading some online sources
that 
Leopard changed a number of things security-wise.  That
said, I 
recommend against making any changes to pwent or its test
until after 
the test can be run again with the public release of Leopard
(to be 
made public in 5 days), in case some relevant detail changed
that the 
change would obscure.

-- Darren Duncan