Pádraig Brady <[email protected]> writes:

>> I hadn't really thought much about the EINTR case, to be honest. I
>> put
>> it there because tee_files() has:
>>      bytes_read = read (STDIN_FILENO, buffer, sizeof buffer);
>>      if (bytes_read < 0 && errno == EINTR)
>>        continue;
>> The 'git blame' says that those lines are from 33 years ago, so I
>> probably shouldn't have placed much weight on it as things may have been
>> different then.
>> I guess they should both be safe to remove, since we don't use any
>> signal handlers. Is that correct?
>
> Yes, that's what I was thinking.
>
> I.e. remove the EINTR handling on the write case anyway
> so that the only functionality change in the patch
> is the stdio removal.

Makes sense, I pushed the patch with the tiny change removing the EINTR
check from write [1].

> Removing EINTR on the read path would be another change,
> which I'm not sure about TBH.

Sure, no problem. We can leave it for now.

Thanks for the review,
Collin

[1] 
https://github.com/coreutils/coreutils/commit/9b855166ed3366f0d335477c2bfa99e5cf9b800f

Reply via email to