…nc (#2545)"

This is what caused regression seen in issue #2943

This reverts commit 30692a2.

Signed-off-by: Adam Treat <[email protected]>

Signed-off-by: Jared Van Bortel <[email protected]>

Signed-off-by: Jared Van Bortel <[email protected]>

layoutChanged is only used when the *order* changes. When rows are
inserted or removed, one of the other three signals is fired.

Signed-off-by: Jared Van Bortel <[email protected]>

Signed-off-by: Jared Van Bortel <[email protected]>

The locking strategy used by ModelList assumes that only one thread will
be trying to update the model list at a time, as writes cannot be
implemented in a fully threadsafe manner.

Signed-off-by: Jared Van Bortel <[email protected]>

Signed-off-by: Jared Van Bortel <[email protected]>

Signed-off-by: Jared Van Bortel <[email protected]>

Also, hint to layoutChanged about the specific way in which the model
was sorted.

Signed-off-by: Jared Van Bortel <[email protected]>

dataChanged will not work correctly if we emit it on an old index after
sorting.

Signed-off-by: Jared Van Bortel <[email protected]>

There were issues with removing clones or models with clones from the
model page, as well as trying to remove a model that no longer exists on
disk, despite being in settings. These should now be resolved.

Signed-off-by: Jared Van Bortel <[email protected]>

This should always be used instead of joining the current model path
setting with the filename, as models may be in subdirectories. But
fixing this correctly is not in scope for this PR.

Signed-off-by: Jared Van Bortel <[email protected]>

removeModel is starting to make sense now.

Signed-off-by: Jared Van Bortel <[email protected]>

- The filename must have a version number, or we will possibly load the
  cache for the wrong version of GPT4All.
- The file should be stored in an appropriate folder for cache, not in the
  settings location.

Signed-off-by: Jared Van Bortel <[email protected]>

Signed-off-by: Jared Van Bortel <[email protected]>