Phar Class decompressFiles() bad in PHP 7.4 and 8.x branches

[kwince]

Description

The following code:

...

<?php
$p = new Phar('/home/gearond/my.phar', FilesystemIterator::SKIP_DOTS | FilesystemIterator::UNIX_PATHS, 'my.phar');
$p['myfile.txt'] = 'four score and seven years ago our fathers.';
$p['myfile2.txt'] = 'four score and severn years ago our mothers.';
foreach ($p as $file) {
    var_dump($file->getFileName());
    var_dump($file->isCompressed());
    var_dump($file->isCompressed(Phar::BZ2));
    var_dump($file->isCompressed(Phar::GZ));
}
var_dump($p->getSupportedCompression() );
var_dump( $p->compressFiles(Phar::GZ) );
foreach ($p as $file) {
    var_dump($file->getFileName());
    var_dump($file->isCompressed());
    var_dump($file->isCompressed(Phar::BZ2));
    var_dump($file->isCompressed(Phar::GZ));
}
var_dump( $p->decompressFiles() );
foreach ($p as $file) {
    var_dump($file->getFileName());
    var_dump($file->isCompressed());
    var_dump($file->isCompressed(Phar::BZ2));
    var_dump($file->isCompressed(Phar::GZ));
}
var_dump( $p->extractTo("/home/gearond/Project/exploration", null, true) );
?>

Resulted in this output:

string(10) "myfile.txt"
bool(false)
bool(false)
bool(false)
string(11) "myfile2.txt"
bool(false)
bool(false)
bool(false)
array(1) {
  [0]=>
  string(2) "GZ"
}
NULL
string(10) "myfile.txt"
bool(true)
bool(false)
bool(true)
string(11) "myfile2.txt"
bool(true)
bool(false)
bool(true)
bool(true)
string(10) "myfile.txt"
bool(false)
bool(false)
bool(false)
string(11) "myfile2.txt"
bool(false)
bool(false)
bool(false)
bool(true)

gearond@gearond-XPS-15-9570:~/Project/exploration$ cat *.txt
K�/-R(N�/JUH�KQ(N-K-�S�LM,*VHL�WI��d��K�/-R(N�/JUH�KQ(N-K�S�LM,*VHL�Wɦ%�d��gearond@gearond-XPS-15-9570:~/Project/exploration$ 

But I expected this output instead:

string(10) "myfile.txt"
bool(false)
bool(false)
bool(false)
string(11) "myfile2.txt"
bool(false)
bool(false)
bool(false)
array(1) {
  [0]=>
  string(2) "GZ"
}
NULL
string(10) "myfile.txt"
bool(true)
bool(false)
bool(true)
string(11) "myfile2.txt"
bool(true)
bool(false)
bool(true)
bool(true)
string(10) "myfile.txt"
bool(false)
bool(false)
bool(false)
string(11) "myfile2.txt"
bool(false)
bool(false)
bool(false)
bool(true)
gearond@gearond-XPS-15-9570:~/Project/exploration$ cat *.txt
cat *.txt
four score and severn years ago our mothers.four score and seven years ago our fathers.gearond@gearond-XPS-15-9570:~/Project/exploration$


### PHP Version

php 8.1.2 and 7.4

### Operating System

ubuntu 20.04 and 22.04

[kwince]

When i switch to PharData class, I can't use the 'phar' extension, but it does not FAIL like it should. Since there is no 'isCompressed()' function it should cause an exception.

However, it fails the same exact way.

[cmb69]

Thanks for reporting! I can reproduce the reported behavior (funnily, the given code is basically the same as the example given for Phar::decompressFiles(), but apparently that never worked properly).

Since there is no 'isCompressed()' function it should cause an exception.

There is PharData::isCompressed(); it is just not documented (like 25 other methods of PharData).

[cmb69]

Hmm, there was a related fix years ago (68f5289), but that likely wasn't the proper fix for this issue. Apparently, there is no provision in phar_flush() to cater to uncompressing files; only compression is handled there.

[kwince]

It's not really apparent to me how the phar class relates to the file in it represents. You can unset an instance in of it and then reopen with a new instance and it complains. I'm going to have to write up something for that, maybe. I feel that it might be related however. I think there's some lazy file flushing that is the root of both problems.

…

On Tue, May 3, 2022, 8:13 AM Christoph M. Becker ***@***.***> wrote: Hmm, there was a related fix years ago (68f5289 <68f5289>), but that likely wasn't the proper fix for this issue. Apparently, there is no provision in phar_flush() to cater to uncompressing files; only compression is handled there <https://github.com/php/php-src/blob/722e9b9fa572628eaae6208b048144fce337c383/ext/phar/phar.c#L2817-L2821> . — Reply to this email directly, view it on GitHub <#8469 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADFKN23FLKI6TX75LK3VKTVIE7CVANCNFSM5UZEK3RQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[cmb69]

Phar(Data)::decompressFiles() is totally broken. While it marks the files as being uncompressed (see phar info), it does not actually store uncompressed files, but rather leaves them as they were. That makes it impossible to ever get the uncompressed contents (unless you do that manually). The same happens for ZIP archives.

Furthermore, Phar(Data)::startBuffering() does not work as advertised; calling ::compressFiles() and ::decompressFiles() will still call phar_flush() and friends right away.

Fixing this requires someone who is actually experienced with the working and the implementation of Phars.