Author vstinner
Recipients martin.panter, serhiy.storchaka, steve.dower, vstinner
Date 2016-04-17.14:50:19
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1460904619.89.0.870616577106.issue26770@psf.upfronthosting.co.za>
In-reply-to
Content
> Another question: should we handle EINTR in ioctl() and fcntl()? But this is a different issue.

See attached test stress.py. I tried it with ioctl() and fcntl() implementations of os.set_inheritable() and I got more than 10,000 signals: no syscalls failed with EINTR.

IMHO EINTR is only used when the syscall waits, for example wait for I/O. I don't think that setting FD_CLOEXEC requires any I/O.

Example:

$ ./python stress.py 
got 16871 signals
History
Date User Action Args
2016-04-17 14:50:19vstinnersetrecipients: + vstinner, martin.panter, serhiy.storchaka, steve.dower
2016-04-17 14:50:19vstinnersetmessageid: <1460904619.89.0.870616577106.issue26770@psf.upfronthosting.co.za>
2016-04-17 14:50:19vstinnerlinkissue26770 messages
2016-04-17 14:50:19vstinnercreate