bug#40407: [PATCH] slow ENCODE_FILE and DECODE_FILE

[Mattias Engdegård]

5 apr. 2020 kl. 17.56 skrev Eli Zaretskii <eliz@gnu.org>:

> Once we do set up default-file-name-coding-system, these macros will
> never return their argument (unless someone forcefully sets the
> encoding to nil, in which case they deserve what they get).  Do you
> agree?

Thank you, and yes, I do agree partly: ENCODE_FILE is the identity for all 
unibyte strings no matter the coding system in use.

However, my point (which I didn't do a very good job explaining) was that if 
either ENCODE_FILE or DECODE_FILE are called with the assumption that they 
return a new string, that is at least a latent bug.

Thus I went through them all once again, and found a few questionable calls 
that I'd like to fix. They rely on Fexpand_file_name returning a new string, 
which may or may not be true now but we would be better without such 
assumptions. (I also stumbled on a potential GC-related bug.) Patch attached!

With these fixed, nothing prevents those two functions from using no-copy 
semantics. I agree this approach is better and safer than going straight for 
code_convert_string_norecord in one pass.

0001-Don-t-rely-on-copying-in-EN-DE-CODE_FILE.patch
Description: Binary data