RenameProcessor#isPreviewUsages() already does that using
UsageViewUtil#reportNonRegularUsages() to determine whether the preview
should be shown.

Because of a number of tricky scenarios related to dynamic features
of Python we used to always display a toolwindow with preview of found
usages for top-level definitions, however it doesn't make much sense
for functions and methods not called explicitly anywhere, such as
tests, or those functions that have just been written.

I added a new method forcesShowPreview(PsiElement, UsageInfo[]) in
RenamePsiElementProcessor that has the same purpose but is called
*after* the usage search has been performed, and thus it doesn't make us
search for usages over again in prepareRenaming().

As guitests module dependencies could have a production scope (not a test scope) we should skip them from library license checking, because guitests modules are not included into distribution.

* Separately process records with and without statuses.
* Always expect record start at the start of the record.
* Return one record or null instead of a list.

Item separator is places in between items, and not before record end.

[git] sequentially parse input lines one by one without storing them in the buffer and without using regular expressions

1. Instead of collecting lines into a StringBuilder, analyze them one by one.
2. Process each line character by character.
3. Expect record start to be at line start and record end -- at line end.
4. When something can not be parsed, instead of magically trying to recover, just cleanup and (try to) continue parsing incoming lines.

The following method deals with record start/end in the commit message, when they are not preceded/followed by a line break.

Code that parses output into a list of records is redundant, records are processed one by one (since b63869fb). So both usages of GitLogOutputSplitter are quite similar and can get a commit record instead of raw output.

Also, since GitLogOutputSplitter does some preliminary parsing to determine records boundaries, it makes sense to move parser call there.

... of schema providers factories: different project-related user data (including project reference) could (and is actually) cached on this file objects, which leads to leaks

(cherry picked from commit e68cd277)

more indicators fixed

x

inspection view: run inspections when view is grouped by file only shows one issue per folder (IDEA-180994)