On 2026-09-25 07:09, Pádraig Brady wrote:

The point wasn't about the first read(),
it was about the blocking in the subsequent read().

Agreed. That kdump output indicates a bug in NetBSD when reading from a pipe 
where EOF was already reported.

My wild guess, without looking at the source code, is that the NetBSD code has 
some provision for named fifos (where a later read could indeed hang) that is 
leaking into the code for pipes (where a later read should return 0 and should 
never hang).



Reply via email to