Release 2.4 issues

George
Administrator

Posts: 4,062

Release 2.4 issues Apr 13, 2021 14:44:14 GMT -5

Quote

Post by George on Apr 13, 2021 14:44:14 GMT -5

Robert:
Wow! That's some trace. Methinks there's a 'hole' where some module is exiting incorrectly, not clearing the trace stack properly.

No, I won't beg off, there's obviously a nasty bug. The fact it crashes at 201 is the clue, the trace table is sized @ 200, so the 'not exiting properly' guess is probably not far off. The fact I see the chain of cmds in the trace twice probably says it might work once, if it doesn't blow the table, but a 2nd try pushes it over the limit.

I'll go check the functions listed, hopefully I can spot the one that's triggering it.

George

[UPDATE]
OK, I can duplicate it, now to find it.
[\UPDATE]

Last Edit: Apr 13, 2021 15:03:17 GMT -5 by George

Stefan
Administrator

Posts: 990

Release 2.4 issues Apr 13, 2021 16:44:40 GMT -5

Quote

Post by Stefan on Apr 13, 2021 16:44:40 GMT -5

Robert

You are not alone.

RE: What is odd is that the displayed dump fills up the whole screen, with no way to get out of it except for pressing Escape
I reported same issue. The 'crash panel' that appears and overflows the screen is scrollable (but I can't recall exactly how I scrolled it. Pretty sure I used the mouse, somehow.) If you can get to the bottom of it, there's a YES/NO question asking if you want data saving or just terminate SPFLite.

RE: warning that the STATE is invalid
I have reported same issue even though nothing has changed the file since last save. Not as frequently as you, but I only have STATE ON for two file types.
I haven't yet figured out what circumstances trigger it. The fact that it gives me the file with all the right line labels and correct include/exclude status for the lines just emphasises that the data in the file is unchanged.

Last Edit: Apr 13, 2021 16:45:32 GMT -5 by Stefan

George
Administrator

Posts: 4,062

Release 2.4 issues Apr 14, 2021 10:50:45 GMT -5

Quote

Post by George on Apr 14, 2021 10:50:45 GMT -5

Robert: Finally tracked down the stupid trace 'leak'. As I thought, it was a missing trace exit from one function. Obviously somewhere in editing the line got dropped. Totally dumb, and never shows up in normal short test sessions, you have to do some repetitive stuff to finally grow the 'leak' into a crash.

The problem of the Crash window being so big obviously won't occur under any normal circumstances. Maybe I'll alter it to just dump the last 'n' entries.

George

Stefan
Administrator

Posts: 990

Release 2.4 issues Apr 14, 2021 15:07:41 GMT -5

Quote

Post by Stefan on Apr 14, 2021 15:07:41 GMT -5

George

The command and module lists really are not much use to the user.
The user really needs little more than the YES/NO to the SAVE prompt and maybe the abend code (although it appears to be C0000005 most, if not all, of the time).
You're right to write all the info the cash log that you create for diagnosis purposes.

So that crash window could be a lot smaller, and independent of the number of trace entries trace entries.