Re: [PATCHv3 03/13] xread_nonblock: add functionality to read from fds nonblockingly

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 22, 2015, 00:02 UTC
Message-ID: <xmqqwpvj7240.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <1442875159-13027-4-git-send-email-sbeller@google.com>

Stefan Beller <sbeller@google.com> writes:

Show 34 quoted lines

> This wrapper just restarts on EINTR, but not on EAGAIN like xread
> does. This gives less guarantees on the actual reading rather than
> on returning fast.
>
> Signed-off-by: Stefan Beller <sbeller@google.com>
> ---
>  git-compat-util.h |  1 +
>  wrapper.c         | 22 ++++++++++++++++++++++
>  2 files changed, 23 insertions(+)
>
> diff --git a/git-compat-util.h b/git-compat-util.h
> index c6d391f..9ccea85 100644
> --- a/git-compat-util.h
> +++ b/git-compat-util.h
> @@ -718,6 +718,7 @@ extern void *xcalloc(size_t nmemb, size_t size);
>  extern void *xmmap(void *start, size_t length, int prot, int flags, int fd, off_t offset);
>  extern void *xmmap_gently(void *start, size_t length, int prot, int flags, int fd, off_t offset);
>  extern ssize_t xread(int fd, void *buf, size_t len);
> +extern ssize_t xread_nonblock(int fd, void *buf, size_t len);
>  extern ssize_t xwrite(int fd, const void *buf, size_t len);
>  extern ssize_t xpread(int fd, void *buf, size_t len, off_t offset);
>  extern int xdup(int fd);
> diff --git a/wrapper.c b/wrapper.c
> index 50267a4..54ce231 100644
> --- a/wrapper.c
> +++ b/wrapper.c
> @@ -223,6 +223,28 @@ ssize_t xread(int fd, void *buf, size_t len)
>  }
>  
>  /*
> + * xread_nonblock() is the same a read(), but it automatically restarts read()
> + * interrupted operations (EINTR). xread_nonblock() DOES NOT GUARANTEE that
> + * "len" bytes is read even if the data is available.
> + */

This comment is somewhat misleading to readers who knew there was xread() and this new one is a variant of it. xread() does not say anything about "if the data is available" (it just blocks and insists reading that much if the data is not ready yet). If we drop the "even if ..." from the end, that confusion would be avoided.

Turning EWOULDBLOCK to EAGAIN is a very useful thing to help the callers, and it deserves to be described in the above comment.

Show 5 quoted lines

> +ssize_t xread_nonblock(int fd, void *buf, size_t len)
> +{
> +	ssize_t nr;
> +	if (len > MAX_IO_SIZE)
> +	    len = MAX_IO_SIZE;

This line is under-indented?

Show 16 quoted lines

> +	while (1) {
> +		nr = read(fd, buf, len);
> +		if (nr < 0) {
> +			if (errno == EINTR)
> +				continue;
> +			if (errno == EWOULDBLOCK)
> +				errno = EAGAIN;
> +		}
> +		return nr;
> +	}
> +}
> +
> +/*
>   * xwrite() is the same a write(), but it automatically restarts write()
>   * operations with a recoverable error (EAGAIN and EINTR). xwrite() DOES NOT
>   * GUARANTEE that "len" bytes is written even if the operation is successful.