Hi all,
This is basically a new thread after what has been discussed for
pg_controldata with its error handling for read():
https://www.postgresql.org/message-id/CABUevEx8ZRV5Ut_FvP2etXiQppx3xVzm7oOaV3AcdHxX81Yt8Q%40mail.gmail.com
While reviewing the core code, I have noticed similar weird error
handling for read(). At the same time, some of those places may use an
incorrect errno, as an error is invoked using an errno which may be
overwritten by another system call. I found a funny one in slru.c,
for which I have added a note in the patch. I don't think that this is
worth addressing with more facility, thoughts are welcome.
Attached is a patch addressing the issues I found.
Thanks,
--
Michael