I'm actually working on file support and am currently fiddling with SOURCE-ID, but I noticed something about BLK. The ANS-2012 standard is clear that QUIT needs to set SOURCE-ID to zero to reset back to the regular user input device, but it doesn't mention BLK anywhere. My understanding is that BLK is checked first, before SOURCE-ID is consulted, to see if the input source is a block. That would seem to indicate that QUIT should set BLK to zero.

I did a quick test on TaliForth2 to generate an error, and it does not set BLK back to zero on an error in a block.

Code:

blk  ? 0  ok
1 load Interpreting a compile-only word
blk ? 1  ok

The same kind of test on gforth seems to show that gforth does set BLK back to zero, but gforth also has exception handling (meaning that it might not head all the way to QUIT).

Code:

blk ? 0  ok
4 load
* a block*:4: Undefined word
>>>asdfhi<<<
blk ? 0  ok

I understand that blocks are kind of left-overs from a previous era, but I find them very useful for accessing I2C flash memories so I would like to support them. Should I modify QUIT to store a 0 in BLK (or is there any reason not to)?

I did a disassembly of a couple of different Forths and the only word that might go awry if BLK <> 0 is ?LOADING. And ?LOADING is only used in the word "-->". If an error is in the first block, then "-->" would not get encountered, but if the BLOCK being LOAD'd is error free and the word "-->" is encountered, the calculation to get the next screen block may be off.

I am not willing to break my version of QUIT to trouble shoot this, so this is just a theory. But you can test this by causing an error so BLK <> 0, then try LOAD'ing a screen that uses "-->" at the end, where the screen being LOAD'd does not have any errors.

This was my line of thinking as well. It's just odd that the standard is very specific about SOURCE-ID but doesn't mention BLK directly. It looks like SOURCE-ID was added in DPANS94 (so you can tell the difference between EVALUATEing a string, file input, and normal input). You're supposed to check BLK first, as a historical artifact from older forths, and then use SOURCE-ID if BLK is 0. Because of that, I think I need to have QUIT put a 0 in BLK just so that SOURCE-ID will be checked properly (only when not loading blocks) in other places.