crash at 1:30am almost every night

[johnsonx42]

for just 32 minutes of recording one show, I've got almost 12,000 of these files. so I guess it's not quite thousands per minute, but many hundreds.

i'll let it go like this for tonight, in case the crash happens, but then I'll turn it off and wait for the updated version that works as you intended.

[johnsonx42]

it didn't crash at 1:30am last night, so for now I've turned off the EIT logging and deleted all the files. I got up to almost 100,000 in 2 hours.

[sub]

Yep - it can result in a huge amount of file writes. As I said, it may hurt performance while it's turned on.

Here is an update the should cause it to overwrite the same files instead of creating a new one each time.

[johnsonx42]

thanks sub, I'll put this one in before tonight's recordings.

I didn't notice any performance hit with the original version that kept overwriting the same single file, so this one that creates one per recording should be just the same. The only reason I was worried about the one that created the thousands of files was that it once it crashed it would have been VERY difficult to navigate in a directory which could have easily reached several million files in a few days, even if the lot of them didn't take much space. Plus last I read there's a limit of 4 million files in an NTFS directory, though that may have changed in recent revisions of NTFS.

edit: oops, my recollection was off, the limit is ~4 Billion, not million. I thought the 4 Billion number was for the whole file system, and that the per-directory limit was much less, but apparently there's no distinction.

[johnsonx42]

no crash yet, but the new version of the EIT logging is working exactly as intended. each new recording creates a single new file which is overwritten continuously with the latest bunch of EIT data. hopefully when the crash occurs the matching EIT-xx.log file will contain the offending data.

[johnsonx42]

finally got one. crash.mdmp and eit logs attached. I'm 99.9% sure it's the eit-121.log, but I tossed in 119 and 120 for good measure.

[johnsonx42]

sub, did you discover anything interesting with this?

[sub]

I was able to reproduce the problem with eit-121.log. This patched fixed it for me.

It related to a EIT entry in your broadcaster's stream that didnt have title for the show.

[johnsonx42]

thanks sub, I'm glad you were able to see the problem with that data. this should cross one more crash type off the list; with this fix I think I'm now down to just one common crash type, which pops up about once a week. I'll have to figure out how to get you some debug data on that one.