Dan McGee wrote:
This patch just makes me cry a little bit...its like taking a sledgehammer to kill a fly, unfortunately. All we wanted to do was keep a symlink around, and we've now regressed to using extensions, which we haven't done since pre 3.1.0.
The extension part is only needed here because gzip/bzip2/xz rely on the correct extension. Probably I should have split the patch into two.
$ touch test && xz test && mv test.xz test.abc && file -i test.abc && xz -d test.abc; echo $? `test.xz' -> `test.abc' test.abc: application/x-xz; charset=binary xz: test.abc: Filename has an unknown suffix, skipping 2
I see three options to address this issue: - do not handle it at all; just throw an error and require a noextract entry - test for the mime header AND the extension - always decompress to stdout and write it to 'file - extension', e.g. bzip2 -dcf test.abc > test -> this will not work if there is no extension -> throws an error; noextract entry required Possibly, option 1 is fine as it might not be the common case that the compressed files use unsuitable extensions.