Demo weights 11 kb, but lasts only 1 second

[Memfis]

I got this strange demo yesterday while recording on pe1_phob.wad. I think it happened because I started rerecording, but haven't closed the ENDOOM screen and as result, both demos was recorded in the same file. Is there a way to separate them? The second one was my best exit so far.

heheeh110.zip

[Never_Again]

A curious case: it does appear that there are two demos inside one LMP file. The first one looks complete and is 60 tics long. The second one follows the demo footer of the first and is missing the header but has the complete footer. Unfortunately, the second demo's data section is truncated: it's missing at least half a tic from the start.

I tried recreating the second demo as a separate file, going on various assumptions - only half a tic is missing, the actions at the start match those of the first - and even several ways of merging the two, all to no avail. The resulting demos all desynced a couple of seconds in at the latest. Most likely, a lot more than half a tic is missing from the start.

If you could tell me the time you got on the tally screen for the second demo, I could get an estimate of how much data is missing. As for the demo itself it's not salvageable, I'm afraid.
The only useful conclusion from this seems to be that recording with Fast Exit (Options -> General -> second page -> Miscellaneous) set to No (or "misc_fastexit 0" in the config) is not safe.

[Memfis]

Thanks for the research, very interesting. The time on intermission screen was 1:10.
Then I'm just going to record a better demo. It wasn't very good anyway. :)

[Never_Again]

Hmm, the usable portion of the second demo is 1:12.40, according to LMPC. Therefore there seems to be more data than there should. That's even more odd than I initially thought. Well, I'm going to try going down from 1:11.0 all the way to ...

ermm, never mind. I overlooked that there are actually three demo footers in this demo, and none of the three data sections are anything even close to 1:10. Better record a new one with Fast Exit set to Yes then. ;)

[Quasar]

What's it opening the files with that could possibly this? "w+" ? It shouldn't be doing that AFAIK.

[Grazza]

There appear to be a number of issues with the extended demo format, including some demo corruption that we have seen a while back. It seems to add some data at the start of the demo in all cases (preventing lmpc from working with the demo), and in some cases this intrusive data seems to be a larger chunk that stops the demo working properly.

Personally, I have turned Extended Demo Format off.

[Quasar]

Grazza said:

There appear to be a number of issues with the extended demo format, including some demo corruption that we have seen a while back. It seems to add some data at the start of the demo in all cases (preventing lmpc from working with the demo), and in some cases this intrusive data seems to be a larger chunk that stops the demo working properly.

Personally, I have turned Extended Demo Format off.

IIRC it was output from stdout.txt was it not? Seems like some file handles are getting jumbled about to me...

[dew]

Grazza said:

Personally, I have turned Extended Demo Format off.

i'm about to do the same if this turns out to be the case. any chance this will get resolved quickly?

[Grazza]

I don't know, but I have made a post in the prboom-plus thread to make sure Andrey is aware of the problem.

[Never_Again]

Grazza said:

including some demo corruption that we have seen a while back. It seems to add some data at the start of the demo in all cases (preventing lmpc from working with the demo)

Extended Demo Format does not add anything to the header or the data section of a demo, all it does is append a demo footer. That LMPC chokes on any demo that has data beyond the quit byte (0x80) is LMPC's limitation. Remove the demo footer with a hex editor and LMPC has no problems with the demo.

I looked into TendaMonster's and Memfis' cases and have not been able to reproduce the weird results they got. Neither could TM when asked, and I doubt Memfis could: I tried recording with Fast Exit off, exiting without closing the ENDPIC and recording again and the resulting demo didn't have multiple footers. In both cases the problem appears to be some weird system disorder and/or user error.

I pointed entryway here, he's looking into it. Barring documented, reproducible evidence that proves that Extended Demo Format is the culprit, I urge everyone not to panic. Extended Demo Format is a Good Thing.

edit: elucidation