Aye, sorry, scratch my hasty that. The syscall() must return a good value. OP problem seems to be in the unpack template. Should try native types instead: "L!L!SZ*". Perl L is 4 bytes...

Yup, the problem was with unpack indeed... I changed the line below from:

my ( $ino, $off, $len, $name ) = unpack( "LLSZ*", $buf );
[download]

To:

my ( $ino, $off, $len, $name ) = unpack( "L!L!SZ*", $buf );
[download]

And the program finally worked as expected:

[me@localhost ~]$ ./test2.pl -d sample/ | head
sample/xaaaaaaaaff
sample/xaaaaaaabdu
sample/xaaaaaaabjw
sample/xaaaaaaaaos
sample/xaaaaaaaalu
sample/xaaaaaaaags
sample/xaaaaaaabci
sample/xaaaaaaabhy
sample/xaaaaaaabjv
sample/xaaaaaaaaxa
[download]

All good, but some questions remains:

1. What could have happened that cause this change in the system? Originally I didn't need to worry about that. Maybe some change related going from 32 to 64 bits platforms?
2. How can I double check if I need to use "native" sizes with the unpack template? Besides the program malfunctions, of course. :-)
Alceu Rodrigues de Freitas Junior
---------------------------------
"You have enemies? Good. That means you've stood up for something, sometime in your life." - Sir Winston Churchill

Oh yes, infinite loop is a problem here, since I'm getting 1 or zero from $len. Here is the updated code:

The part changed to Pod is exactly the one that put me in a "infinite" loop (probably not infinite, but long enough to create a very large file, but I killed the process once it reached GB).

As you can see in the code, I used the same buffer size as in the C code... but results are the same: