Nice! Yes I saw the console.log() haha, but anyhow it looks it sufficient ;)
Thanks for sharing, but I wonder why this happened now...

I left a copy of the original preg above the update in the script plus a backup script appended with _ in the same dir :)

I used sample values returned for [c]$post_data[$col_msg_text][/c], extracted/substituted values/variables for [c]$post_data[$col_bbcode_uid][/c]
and tested the original capturing sub-pattern matches in [c]$code_data[ 1 ][ 0 ][/c] and [c]$code_data[ 2 ][ 0 ][/c].
Under the existing pattern, [c]$code_data[ 2 ][ 0 ][/c] didn't return anything ( when the code blocks are named as lexikos stated ), which was the issue.

I just simplified the approach of obtaining the capturing sub-pattern matches and stress tested it.
(wish there was an easier/cleaner way to debug though)

// preg updated by TLM
	preg_match_all(	"#\s+file=(.*?):"
					. $post_data[$col_bbcode_uid] . "\](.*?)\[/code:"
					. $post_data[$col_bbcode_uid] . "\]#msi"
					, $post_data[$col_msg_text]
					, $code_data);

great! what was the fix in general?
it would be great to know for future purposes. it happened to me once where i needed to know what i did before and luckily i had it posted somewhere.. ;)

[b]Edit[/b] i think it was simply the regex going bezerk?
[code=php file=test.php]// preg updated by TLM
preg_match_all( "#\s+file=(.*?):"
. $post_data[$col_bbcode_uid] . "\](.*?)\[/code:"
. $post_data[$col_bbcode_uid] . "\]#msi"
, $post_data[$col_msg_text]
, $code_data);[/code]

Simplified some unnecessarily complicated preg.
Seems to be working.

Looks like TLM is working on it right now ;)

Test2

Test

It appears that when there are multiple code blocks, [strike]the filename of the first one (if any) is used[/strike] the download link downloads the first code block, even if that one doesn't have a download link. If I click the download link below, I get Untitled.txt (Test) on Chrome 42 and IE11:

[code=autohotkey]Test2[/code][code=autohotkey file=titled.ahk]Test[/code]

Manually adjusting the URL from part=1 to part=2 gives me the correct code block in this case.