By default, git apply isn't very good at handling patches that don't

apply perfectly cleanly. It will just fail with an error. That means if

you attempt to apply a patch, and the underlying files have changed,

even just a little, nothing will be applied, and the user will be shown

an error.
git apply has a nice option we can use to force better behavior:

-3 (or --3way). This enables Git's merge behavior for patches, and will

also generate conflict markers. This option implies --index, so that has

been removed from the command line.
When a patch does fail to apply, partially or fully, it no longer shows

an "Unable to execute command" error. Instead, apply_patch() returns a

PatchResult, which contains the state of the patch operation, including

any list of conflicting files that could be reported, and any output

from the patch command line.
This is then shown to the user. The output shows that the patch is being

applied, then shows the patch output, and then the status as a

human-readable message. If any conflicting files were seen (Git only,

for now), those will be shown as well.

Tested with both Git and Mercurial, in all failure/success modes.
For Git, I tested with and without a list of conflicting files.