wp_convert_bytes_to_hr tests and NaN

[soulseekah]

[1218/tests] were written for the deprecated wp_convert_bytes_to_hr, these tests are broken at least on 64-bit machines, at most - on all machines.

On 64-bit PHP (PHP_INT_SIZE == 8) floating point does not match expected output (gets 1022.9990234375MB). The last digits in floating point should not be depended upon due to these inaccuracies. Rounding in PHP < 5.3 is not an option either, since rounding mode was introduced later and different systems get different results (round up vs. round down by default). So perhaps looking at error and discrepancy would be a decent solution (maybe even less error 0.00001?).

Next [23440], in the function itself we get (int)double(NAN) result from feeding -1 and 0, which results in $power to be a huge negative number instead of 0. This results in both a Warning (undefined array index) being thrown, and the "B" missing in tests, i.e. expected "NANB" != "NAN" and expected "0B" != "0".

Related #19067

[SergeyBiryukov]

#12915 suggested rounding the output to 2 decimal digits.

#22295 also mentioned the notice.

[soulseekah]

See: ​http://www.php.net/manual/en/function.round.php#98704 for PHP < 5.3 round direction is not predictable it seems.

[SergeyBiryukov]

In 1220/tests:

Fix wp_convert_bytes_to_hr() test for 64-bit systems. props soulseekah. see #23626.

[SergeyBiryukov]

In 23502:

Avoid an undefined offset notice in wp_convert_bytes_to_hr(). props soulseekah. fixes #23626.

[SergeyBiryukov]

Still getting different results:

• 32-bit, wp_convert_bytes_to_hr( -1 ): NANB
• 64-bit, wp_convert_bytes_to_hr( -1 ): -1B

I didn't commit 19067.function.patch​ as is, because for 1 terabyte, the function would return "1B" (before [23502] it would be "1" and a notice).

Testing 23626.patch​ and 23626.test.patch​ now.

[soulseekah]

Perhaps it should return false for all sorts of weirdness? Or 0B.

[SergeyBiryukov]

Replying to soulseekah:

Perhaps it should return false for all sorts of weirdness? Or 0B.

It's supposed to return a string. It also does not state that $bytes should be positive, so for "-1", I guess "-1B" is a more appropriate result than "0B".

We could also add "TB" to the array: 23626.2.patch​, 23626.test.2.patch​.

So far getting consistent results with the latest patches.

[SergeyBiryukov]

In 1226/tests:

Add more assertions for wp_convert_bytes_to_hr(). Correct expected result for a negative number. see #23626.

[SergeyBiryukov]

In 23551:

Make wp_convert_bytes_to_hr() return consistent results on 32-bit and 64-bit systems. fixes #23626.

[SergeyBiryukov]

In 1270/tests:

Fix wp_convert_bytes_to_hr() test for decimal comma locales. see #23626.

[SergeyBiryukov]

In 39265:

Tests: Use assertEquals()' native functionality for delta comparison in test_wp_convert_bytes_to_hr().

See #23626.

[SergeyBiryukov]

In 39267:

Tests: Add a missing $message argument for assertEquals() in [39265].

See #23626.