bug#44007: 26.3; Strange shell mode performance

[Dmitry Gutov]

On 06/11/2023 15:37, Herman@debbugs.gnu.org wrote:
> I think it is a bug because Emacs doesn't try to read again if one 
> read() was successful.  My hacky solution fixes that.  With the extra 
> cost of one read syscall.  But we can do better: we can avoid it if the 
> first read call already read 'nbyte' bytes, then there is no need to 
> call another read().

FWIW, when I tested different options for bug#66020 (see for example 
https://debbugs.gnu.org/66020#58), increasing the default buffer size is 
anyway the simplest way to get to top performance without sacrificing 
reliability.

There are some patches attached in the subsequent message, BTW, that one 
could try out.