Previously, we only added the new directory and not the contents. The new AssetDataGatherer has this assumption now though:
// Note that we AddDiscovered but do not scan the directory
// Any files and paths under it will be added by their own event from the directory watcher, so a scan is unnecessary.
// The directory may also be scanned in the future because a parent directory is still yet pending to scan,
// we do not try to prevent that wasteful rescan because this is a rare event and it does not cause a behavior problem
When we iterate the contents of a new directory we now add them to the FileChanges array to match the above.
[at]Brandon.Schaefer, [at]Will.Damon
#jira UE-131552
#ROBOMERGE-AUTHOR: michael.sartain
#ROBOMERGE-SOURCE: CL 17907948 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v883-17842818)
[CL 17907958 by michael sartain in ue5-release-engine-test branch]
We still have some duplicate inotify watches, but this first pass will spew a lot more information when we hit inotify limits.
Adds a "DumpINotifyStats" command in non-release builds
Spews global inotify & UE stats, along with physical count of directories, etc.
Canonicalize directory path in FDirectoryWatcherLinux::RegisterDirectoryChangedCallback_Handle
Shootergame was adding 141 duplicate watches for Samples/Games/ShooterGame/Content w/o this. Was 1136, is now 995.
Change PathsToWatchDescriptors tmap to PathNameHashSet
Don't need to store full paths for each watch directory twice
Fix bugs in TestPAL in addition to adding DumpStats() command, which looks ~ like this:
LogDirectoryWatcher: Warning: inotify limits
LogDirectoryWatcher: Warning: max_queued_events: 16384
LogDirectoryWatcher: Warning: max_user_instances: 128
LogDirectoryWatcher: Warning: max_user_watches: 65536
LogDirectoryWatcher: Warning: inotify per-process stats
LogDirectoryWatcher: Warning: systemd (pid 2239) watches:23 instances:3
...
LogDirectoryWatcher: Warning: plugin_host-3.3 (pid 395041) watches:62 instances:1
LogDirectoryWatcher: Warning: plugin_host-3.8 (pid 395044) watches:62 instances:1
LogDirectoryWatcher: Warning: TestPAL (pid 396852) watches:2 instances:1
LogDirectoryWatcher: Warning: Total inotify Watches:392 Instances:28
LogDirectoryWatcher: Warning: Current watch requests
LogDirectoryWatcher: Warning: /var/tmp/DirectoryWatcherTest396852: 2 watches
LogDirectoryWatcher: Warning: Total count:2
The above is also dumped (once) when we fail to init or add a inotify watch.
Need to create better documentation and add a pointer to it, similar to what VSCode does: (hat tip Brandon)
https://code.visualstudio.com/docs/setup/linux#_visual-studio-code-is-unable-to-watch-for-file-changes-in-this-large-workspace-error-enospc
Related bugs:
; FPS BP Cooking Content - errno = 28, Out of inotify watches
https://jira.it.epicgames.com/browse/UE-125210
; inotify Warnings when Cooking Content for Linux
https://jira.it.epicgames.com/browse/UE-119696
; Time Niagara Sequencer failed to play | Error: Couldn't find file for package
https://jira.it.epicgames.com/browse/UE-89750
; inotify warnings from Linux command line builds
https://jira.it.epicgames.com/browse/UE-76562
[at]Brandon.Schaefer, [at]James.Singer
#jira UE-76562, UE-89750, UE-119696, UE-125210
#ROBOMERGE-AUTHOR: michael.sartain
#ROBOMERGE-SOURCE: CL 17498916 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v870-17433530)
[CL 17498920 by michael sartain in ue5-release-engine-test branch]
ANSI_TO_TCHAR should be UTF8_TO_TCHAR for event name
[FYI] Brandon.Schaefer
#jira UE-50360
#rb none
#ROBOMERGE-AUTHOR: michael.sartain
#ROBOMERGE-SOURCE: CL 17452991 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v870-17433530)
[CL 17452999 by michael sartain in ue5-release-engine-test branch]
#rb trivial
Removed references to UE4 in comments
#ROBOMERGE-SOURCE: CL 15796780 in //UE5/Release-5.0-EarlyAccess/...
#ROBOMERGE-BOT: STARSHIP (Release-5.0-EarlyAccess -> Main) (v783-15756269)
[CL 15797094 by mark lintott in ue5-main branch]
#rb arciel.rekman
#ROBOMERGE-SOURCE: CL 12096923 via CL 12096925 via CL 12096927
#ROBOMERGE-BOT: (v657-12064184)
[CL 12096928 by graeme thornton in Main branch]
#rnx
#rb none
#ROBOMERGE-SOURCE: CL 10869240 via CL 10869516 via CL 10869902
#ROBOMERGE-BOT: (v613-10869866)
[CL 10870584 by ryan durand in Main branch]
Get GIsRequestingExit now by IsEngineRequestingExit()
Set GIsRequestingExit now by RequestEngineExit(const TCHAR* Reason) or RequestEngineExit(const FString& Reason)
NOTE If Reason is 4 or less chars it will generate an ensure to force a reason to exit
The reason behind this change is right now setting GIsRequestingExit to true can cause many things to break mainly early on and with out any sort of log warning we have entered this state. We should wrap this behind a function to allow for proper handling
#rb Chris.Babcock, Michael.Trepka, Michael.Noland
#jira UE-79933
[FYI] Michael.Noland
#ROBOMERGE-SOURCE: CL 8649683 via CL 8653683
#ROBOMERGE-BOT: (v417-8656536)
[CL 8658680 by brandon schaefer in Main branch]
- Errors cause Cook commandlet to exit with 1, while this error is arguably benign.
#rb none
[FYI] Brandon.Schaefer
#jira none
#ROBOMERGE-SOURCE: CL 5159901 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 5207560 by arciel rekman in Main branch]
