On Wed, 10 Nov 2010 13:25:56 +0000 Nick Treleaven nick.treleaven@btinternet.com wrote:
On Tue, 9 Nov 2010 21:27:07 +0200 Dimitar Zhekov dimitar.zhekov@gmail.com wrote:
Well, g_file_set_contents() uses buffered I/O, so obviously it's acceptable too.
Yes, it uses fwrite and then errno.
Probably fwrite(), fflush() and then errno? fflush() is guaranteed to set errno AFAIK.
No:
errno = 0; n_written = fwrite (contents, 1, length, file); if (n_written < length){ save_errno = errno;
g_set_error (err, G_FILE_ERROR, g_file_error_from_errno (save_errno), _("Failed to write file '%s': fwrite() failed: %s"), display_name, g_strerror (save_errno));
fclose (file);
Ah, it sets errno = 0 first, assuming that fwrite() may fail to do so.
I've now committed a fix so that any write failure should be reported to the user even if errno is 0. Please test.
It won't be 0 unless you assign it. A successful library call does not clear errno.
Why fake 0 errno as EIO? g_file_set_contents() doesn't, and *strerror() returns something for 0, usually "Error 0".
fclose check is fixed now too.
If fwrite() and fclose() both fail, I'd prefer to see the errno of fwrite(), even if 0 (g_file_set_contents() does that). Hmmm, it'll be nice to display the filename as UTF-8 too, plus the name of the failed function, and we have a GError ready...
Mind if I try to improve the error handling, using gfileutils.c as a source? There will be 3 new translation strings.