Everything

php

new bug in 5.1.0b3 with length param in fread() with local files

php.internals

Greg Beaver

21 years ago

Hi,

I have noticed that fread()'s behavior has changed between PHP 5.0.4 and
PHP 5.1.0b3.  Basically, http://pear.chiaraquartet.net/go-pear.phar
works in PHP 5.0.4 and fails in 5.1.0b3 because of this change.  The
line in question of the .phar from PHP_Archive is:

fread($this->_file, $this->internalFileLength +
$this->footerLength);

Basically, a .phar is a .tar.  For files inside the .phar larger than
16000 bytes, the fread above fails with the incorrect error message:

PHP Warning:  fread(): Length parameter must be greater than 0 in
c:\php5\pear\PHP\Archive.php on line 193

Note that the failure is OS-independent.

However, the actual length parameter is 21,155, which is not less than
zero, unless you consider it to be a negative number in binary.  After
the fread in PHP 5.0.4, ftell() reveals that the pointer has advanced to
22016, whereas in 5.1.0b3, it has only advanced to 16384.

According to the notes in http://bugs.php.net/bug.php?id=30936 this
should work with a local file, so should I re-open 30936, or is this a
new bug?

Greg

Wez Furlong

21 years ago

If you've spotted a behaviour change, open a new bug with a short
self-contained reproducing script.  I suspect some kind of engine bug
here, because what you've described doesn't make sense :)

That warning about the length comes from this code in the fread function:

len = Z_LVAL_PP(arg2);
if (len <= 0) {
php_error_docref(NULL TSRMLS_CC, E_WARNING, "Length parameter
must be greater than 0");
RETURN_FALSE;
}

So either you are passing in something that evaluates as 0 (or less),
or something is corrupting memory and breaking the length value by the
time it reaches fread().

I'd recommend running your script under valgrind
(--disable-memory-manager when you configure PHP) and see if it
reports anything bad.

--Wez.

On 7/22/05, Greg Beaver <cellog@php.net> wrote:

Greg Beaver

21 years ago

Wez Furlong wrote:

> If you've spotted a behaviour change, open a new bug with a short
> self-contained reproducing script.  I suspect some kind of engine bug
> here, because what you've described doesn't make sense :)
>
> That warning about the length comes from this code in the fread function:
>
>     len = Z_LVAL_PP(arg2);
>     if (len <= 0) {
>         php_error_docref(NULL TSRMLS_CC, E_WARNING, "Length parameter
> must be greater than 0");
>         RETURN_FALSE;
>     }
>
> So either you are passing in something that evaluates as 0 (or less),
> or something is corrupting memory and breaking the length value by the
> time it reaches fread().

In the attempt to find a short reproducing test script, I determined
that the warning was caused by an earlier iteration through the method
that uses fread.  So, the warning was valid (0 was being passed in).
What this means, however, is that the script is in fact simply failing
to read in the requested amount with no warning at all.  I'll have a
test script shortly for the bug report.

Greg

Andi Gutmans

21 years ago

Has anyone had a chance to look at this problem?

At 10:08 PM 7/21/2005 -0600, Greg Beaver wrote:

Wez Furlong

21 years ago

From the description, it doesn't sound like a bug.
Awaiting more details.

--Wez.

On 7/25/05, Andi Gutmans <andi@zend.com> wrote: