Request for comment: Character substring lengths

[Tobias Schlüter]

Brooks Moses wrote:
> Tobias Schlüter wrote:
>> FX Coudert wrote:
>>  > I think there might be lots of places in the front-end relying on this
>>  > convention. A grep for "ts.cl " reveals a few trivial ones, but I
>>  > think there might be more...
>>
>> Indeed, I found about this recently when I prepared my last series of 
>> patches.  I think I wrote something about this in one of my patch 
>> submissions, but I couldn't find it, so I may only have it in my 
>> private notes.  I will check later today if I kept them, as I then 
>> believed to have understood the mechanism behind charlenghths.
> 
> If you find something, could you put it in gfc_internals.texi?  :)

It was the patch for PR31266 that made me look at these things.  There's 
not much in my notes.  Basically, ts.cl == NULL means length unknown. 
This of course bears the question if ts.cl == NULL implies ts.cl->length 
== NULL, in which case the double checks for the negative of both 
conditions would be redundant.

> There's also the complication that constant character expressions don't 
> have ts.cl.  (Or, at least, they don't reliably have it; I'm not at all 
> sure of the details.)  Instead, they have value.character.length.

I think the assumption is that the length of an EXPR_CONSTANT is always 
expr->value.character.length, and no gfc_charlen is needed for these. 
Whether that is the best way to organize this, I don't know.

- Tobi