File-creation mode should respect umask

[micahjsmith]

I think that file creation using open should respect the mask set by umask (at least on *nix). Currently, open calls ios_file which uses a file-creation mode given in src/support/ios.c:

fd = open_cloexec(fname, flags, S_IRUSR | S_IWUSR /* 0600 */ | S_IRGRP | S_IROTH /* 0644 */);

This gives every file user read-write permissions and group and other read-only permissions, regardless of the value of umask. On a multi-user system, teams may use the umask builtin to specify a mask for file-creation; it makes sense to me to respect this in open.

[vtjnash]

can you produce an example of what you want and when it fails? You can't ignore umask since it is enforced by open:

http://linux.die.net/man/2/open

mode specifies the permissions to use in case a new file is created. This argument must be supplied when O_CREAT is specified in flags; if O_CREAT is not specified, then mode is ignored. The effective permissions are modified by the process's umask in the usual way: The permissions of the created file are (mode & ~umask). Note that this mode only applies to future accesses of the newly created file; the open() call that creates a read-only file may well return a read/write file descriptor.

[micahjsmith]

Sure. I mean that either there should be a way to specify exactly the file-creation mode (I don't think any of the existing open methods allow that) or exact inverse of umask could be used. For example

shell> umask 0002
julia> open("test.txt", "w") do f
       println(f, "hello")
       end
shell> echo "hello" > test1.txt
shell> ls -l test.txt test1.txt
-rw-rw-r-- 1 user group 6 May  6 15:48 test1.txt
-rw-r--r-- 1 user group 6 May  6 15:47 test.txt

So I'm not suggesting that there is any problem with umask/open failing but that in the absence of an ability to control file-creation mode it makes sense to use umask inverse as default (as echo seems to do).

And I think that even if there were an open method allowing this behavior, so many other functions like writetable rely upon the normal version of open that an environment variable/OS setting may make sense.

[micahjsmith]

To follow up, given that file creation mode is given by mode & ~umask, I wonder why Julia's open doesn't use the same default mode as Linux's open, which seems to be read and write for user, group, and other. I realize that my original issue doesn't make too much sense, so I'm closing and will open a new issue.