git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v6 5/6] Change copy_fd() to not close input fd

From
Junio C Hamano <gitster@pobox.com>
Date
Aug 26, 2014, 17:10 UTC
Message-ID
<xmqqd2bnjhx1.fsf@gitster.dls.corp.google.com>
In-Reply-To
<1409066605-4851-6-git-send-email-prohaska@zib.de>
Steffen Prohaska <prohaska@zib.de> writes:
> The caller opened the fd, so it should be responsible for closing it.

Hmmmm, I am not sure what the benefit of such a dogmatism. This function consumes all that is useful in the fd, and there is nothing interesting that can be do further on it, no?

Ah, wait. The caller could choose to rewind and reuse the contents, and it is very selfish of this function to unilaterally declare that once you give an fd to it you cannot do anything with it laster.

So I think this is a good change, but the justification is not quite. It is not "the caller should be responsible"; it is more about "the callee did not open it, does not own it, and should allow the caller, if it chooses, reuse it by seeking after the callee is done."

>> Subject: Re: [PATCH v6 5/6] Change copy_fd() to not close input fd
Let's follow the "<area>: <description>" convention here, too, e.g.
    copy_fd(): allow callers to work on input fd after it returns
or something.
Thanks.
Show 58 quoted lines
>  copy.c     | 5 +----
>  lockfile.c | 3 +++
>  2 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/copy.c b/copy.c
> index a7f58fd..d0a1d82 100644
> --- a/copy.c
> +++ b/copy.c
> @@ -10,7 +10,6 @@ int copy_fd(int ifd, int ofd)
>  			break;
>  		if (len < 0) {
>  			int read_error = errno;
> -			close(ifd);
>  			return error("copy-fd: read returned %s",
>  				     strerror(read_error));
>  		}
> @@ -21,17 +20,14 @@ int copy_fd(int ifd, int ofd)
>  				len -= written;
>  			}
>  			else if (!written) {
> -				close(ifd);
>  				return error("copy-fd: write returned 0");
>  			} else {
>  				int write_error = errno;
> -				close(ifd);
>  				return error("copy-fd: write returned %s",
>  					     strerror(write_error));
>  			}
>  		}
>  	}
> -	close(ifd);
>  	return 0;
>  }
>  
> @@ -60,6 +56,7 @@ int copy_file(const char *dst, const char *src, int mode)
>  		return fdo;
>  	}
>  	status = copy_fd(fdi, fdo);
> +	close(fdi);
>  	if (close(fdo) != 0)
>  		return error("%s: close error: %s", dst, strerror(errno));
>  
> diff --git a/lockfile.c b/lockfile.c
> index 2564a7f..2448d30 100644
> --- a/lockfile.c
> +++ b/lockfile.c
> @@ -224,8 +224,11 @@ int hold_lock_file_for_append(struct lock_file *lk, const char *path, int flags)
>  	} else if (copy_fd(orig_fd, fd)) {
>  		if (flags & LOCK_DIE_ON_ERROR)
>  			exit(128);
> +		close(orig_fd);
>  		close(fd);
>  		return -1;
> +	} else {
> +		close(orig_fd);
>  	}
>  	return fd;
>  }
Previous: Steffen ProhaskaNext: Jeff King
Message 15 of 19 in “Stream fd to clean filter; GIT_MMAP_LIMIT, GIT_ALLOC_LIMIT with git_env_ulong()”
  1. 0/6 Stream fd to clean filter; GIT_MMAP_LIMIT, GIT_ALLOC_LIMIT with git_env_ulong()Steffen Prohaska, Aug 26, 2014
  2. 1/6 convert: drop arguments other than 'path' from would_convert_to_git()Steffen Prohaska, Aug 26, 2014
  3. 2/6 Add git_env_ulong() to parse environment variableSteffen Prohaska, Aug 26, 2014
  4. Jeff KingAug 26, 2014
  5. Junio C HamanoAug 26, 2014
  6. Jeff KingAug 26, 2014
  7. Junio C HamanoAug 26, 2014
  8. Jeff KingAug 27, 2014
  9. Junio C HamanoAug 27, 2014
  10. Steffen ProhaskaAug 28, 2014
  11. Junio C HamanoAug 28, 2014
  12. 3/6 Change GIT_ALLOC_LIMIT check to use git_env_ulong()Steffen Prohaska, Aug 26, 2014
  13. 4/6 Introduce GIT_MMAP_LIMIT to allow testing expected mmap sizeSteffen Prohaska, Aug 26, 2014
  14. 5/6 Change copy_fd() to not close input fdSteffen Prohaska, Aug 26, 2014
  15. Junio C HamanoAug 26, 2014
  16. Jeff KingAug 26, 2014
  17. Steffen ProhaskaAug 28, 2014
  18. Junio C HamanoAug 28, 2014
  19. 6/6 convert: stream from fd to required clean filter to reduce used address spaceSteffen Prohaska, Aug 26, 2014

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.