[mimeset] silently fails on symlinks

[diver]

This is hrev44503.

mimeset -f /boot/home/Welcome

doesn't identify Welcome link on the Desktop and doesn't print anything. Identify/Force identify from Tracker doesn't work either.

[anevilyak]

It turns out that it is actually succeeding ; at identifying the link itself. The registrar opens the supplied ref such that it's not followed if it's a link. As a consequence, the sniffer is only run on the link itself, which obviously just gets identified as a symlink. I'd propose that if the ref upon which a mime update is requested is a link, then we also update info for its target, since that's the more likely behavior anticipated by the user. Will leave the ticket open for now in case anyone objects/has an alternative suggestion, otherwise I have a trivial patch which would handle the above change.

[leavengood]

I think the proposed fix makes sense. I can't imagine there ever being a time when the user really wants to make sure something is a symlink by identifying it. And is a symlink ever not already identified as a symlink?

[bonefish]

The update_mime_info() and create_app_meta_mime() (and mimeset) behavior matches BeOS'. I also think it makes sense, since those functions have a recursive mode and you definitely don't want to follow symlinks in that mode, since that could result in infinite recursion or wild jumps.

I don't think the behavior of the functions or mimeset should change. If you want an initial symlink to be resolved, it is rather simple to do that before calling the function/tool, using the BeAPI (e.g. BEntry) or in the shell readlink.

I agree that Tracker's "Identify" feature should be more intelligent, though.

[anevilyak]

Understood, will tweak Tracker then.

[diver]

Unfortunately this won't help ​mime_update.sh script to resolve Welcome and User Guide symlinks.

[anevilyak]

That could alternatively be fixed by adding:

mimeset       -f "$SYSTEM/bin"

To the list of things that script touches.

[diver]

Yes, but this will take even more time on the first boot from ISO.

[bonefish]

Please note that the whole mime_update.sh thing is a temporary work-around (since the build system should do that already). So I wouldn't try to optimize all that much. Instead of adding the whole system bin folder only the two concerned files could be processed, though.

[anevilyak]

Fixed in hrev44513.