wirehead wrote: ↑July 22nd, 2022, 3:27 pm
GUYTU6J wrote: ↑July 21st, 2022, 10:14 pm
Interestingly, the crash for 2 and 26 is expected but 27 is not. Does the message I get for the first two match yours?
Code: Select all
Fatal error: Maximum execution time of 90 seconds exceeded in /home/psstoref/conwaylife/forums/includes/utf/utf_tools.php on line 107
Yes, that's the same one I get for 27. I've never seen that error before -- it's not the same one as in the OP, it's just a cryptic line of unformatted HTML. Weird.
So my previous idea could be a bit misleading... The two errors, "MySQL gone away" and "execution time exceeded" respectively, may have distinct and unrelated causes. Can you reproduce the "MySQL gone away" error with some other [y] in close proximity in the [keyword1] test?
...
Without: Works, and the only result is
this post.
Strange...
What does all this mean? Is it a bug in how Nathaniel has it set up, or, is it a bug in phpBB itself?
These two results match my anticipation well. (For those who want to repeat these tests, note that due to LaundryPizza03's two posts in the OCA subforum today containing [keyword2], the [x] above is displaced by 2.) That hyperlinked sole result has a hundred-line Golly script using a lengthy base64 encoding in
one line (i.e. no linebreaks).
As far as I see from outside, when the post is picked up by a query in some mode (explained below), the utf_tools.php will attempt to locate the input, truncate other characters nearby and highlight the keyword in the displayed list. However, processing a one-line base64 encoding from a long post accordingly will take overwhelmingly long, and printing the very long <s>boat</s> post to the screen can also be memory-intensive. Eventually it gives up, providing an "execution time exceeded" error.
Generally every piece of overly long
one-line content could be a potential threat, not limited to random base64 encodings, as exemplified by
the top post from this query:
Code: Select all
https://conwaylife.com/forums/search.php?keywords=[keyword1]&start=291
The search mode, determined by url parameters, is a key factor in performance. In this case, the "&t=5588" search mode shows the post as a whole, preserving the code box format. I suspect that in other modes, the code box format is disrupted, which contributes to the "execution time exceeded" error as well. In the last comparison, the "&ch=15" mode for truncating other characters around the keyword to 15 is believed to disrupte the code box.
All of the primary analysis above is from an Amateur Crash Investigator; where are the truly responsible ones for pan-pan signals like this?