GitLab

refs #268

Fix #265

Don't delete a folder stored in a removable device if the root folder
can't be read. This is likely to be because the device was just mounted
and is not ready to be read from.
This is preventing any folder I indexed from my NAS to be removed when I
power it on:
MediaLibrary.cpp:2715 onDeviceMounted Device doop changed presence state: 0 -> 1
[...]
DiscovererWorker.cpp:527 runReloadDevice Reloading entrypoint on mounted device: smb://192.168.0.101:445/.../
FsDiscoverer.cpp:355 checkFolder Failed to browse smb://192.168.0.101:445/.../: Failed to browse network directory: Unknown error: Input/output error
FsDiscoverer.cpp:172 checkRemovedDevices The device containing smb://192.168.0.101:445/.../ is removable. Refreshing device cache...
[...]
MediaLibrary.cpp:870 deleteFolder deleting folder smb://192.168.0.101:445/.../

So we can have a different strategy when discovering a root folder or a
subfolder

Fix #264

Since it contains the latest dvdread/dvdnav release, which will avoid
the logs being spammed

The thread wrapper doesn't handle the threads we create there

As it was invoked in the ostream operator<< overload in case no standard
overload was available, causing an infinite recursion

rvalue reference to a string were passed to sqlite as a static value,
while its lifetime ends as soon as the function returns.
This also removes support for non-const char*, as we don't expect those.
We now pass const char* (hardcoded strings) and const std::string& using
SQLITE_STATIC, and make sqlite copy temporary buffers (std::string&&)
using SQLITE_TRANSIENT
This fixes a use after free/stack-use-after-scope when passing a
std::string&& to a request helper

Since we can now restore a file from its mrl before the device lister
marks the device back as present

This works around false negative when a file was added from an host, and
fetched with another equivalent one.

Fix #263

Except for logging purposes, there was no point in fetching any FS
related entity. This is done when fetching the folder, so this work was
duplicated here.

This reverts commit a28df82e.
This was a very limited hack since the device are not guaranteed to be
available as soon as the device lister is started.
Revert it before implementing a more efficient solution for early
devices fetching

refs #263

This way we'll be able to update it in DB without fetching the known
mountpoints first, and guessing which one was added/removed
refs #263

refs #263