lwjgl2 optionally requires the xrandr command line utility on linux, but
does not check if the executable actually exists before trying to use
it. We can force it to fall back to the xf86videomode implementation
by checking for the executable ourselves, and force disabling xrandr
with this boolean [1] if it does not exist.
Link: 2df01dd762/src/java/org/lwjgl/opengl/LinuxDisplay.java (L214) [1]
Signed-off-by: Blake Batson <bbatson101@gmail.com>
(cherry picked from commit e0c2fbbcde)
fixes#4537
This ensures that when looking to uninstall a resource prism will
consider the disabled ones to.
Right now we have a guard in place to prevent resources using the same
name so this check will allways match with the correct resouce.
Signed-off-by: Trial97 <alexandru.tripon97@gmail.com>
(cherry picked from commit 08c45684de)
introduced here https://github.com/PrismLauncher/PrismLauncher/pull/3260fixes#4415
reason: some snapshot have Pre-Release in our meta but when searching in
Modrinth this needs to be translated to -pre and the reverse needed to
be done for filtering after we fetched the version.
Now there are snapshots with -pre in name and that works with Modrinth
but when we translate it back we replace it with Pre-Release so the
easeiest patch is just to double the version(one with -pre one with
Pre-Release)
The correct one would be to complicate the code and identify the
versions that need the transition and only apply this for those.
Signed-off-by: Trial97 <alexandru.tripon97@gmail.com>
(cherry picked from commit b1408775b3)
As it turns out, LLDB can load Windows crash dumps!
This allows us go back to the regular (better supported) DWARF debuginfo
format used by MinGW, as now we have a tool that can both parse those
symbols *and* Windows' crash dumps. The biggest advantage here is that
once again, MinGW crash dumps can be easily inspected on Linux
Signed-off-by: Seth Flynn <getchoo@tuta.io>
(cherry picked from commit d85ff94f03)
Since we've started using ninja's multi-config generator, evaluating
CMAKE_BUILD_TYPE at configure-time is no longer reliable. Thankfully,
CMake offers "generator expressions" that are evaluated during build
system generation, which allows us to continue using these conditional
flags without much headache
Signed-off-by: Seth Flynn <getchoo@tuta.io>
(cherry picked from commit 85849fea41)
fixes#4547
introduced by #3828
made a mistake to use addonId instead ov fileId
this only applies to modrinth
on curseforge I could not reproduce this bug
Signed-off-by: Trial97 <alexandru.tripon97@gmail.com>
(cherry picked from commit d7745d97f2)
fixes#4541
introduced in
3c46d8a412
The original commit introduced m_name but never used it.
When the endActivity would be called with a count of 0 this would crash
the laucnher.
How to reproduce: try to switch your skin in quick succession until you
get 429 too many requests as response to the login.
I will also presume this is not the only crash caused by this(hopefully
it is as it was not catched for four years)
Signed-off-by: Trial97 <alexandru.tripon97@gmail.com>
(cherry picked from commit f4b22dae90)
This should fix issues with OpenGL, as well as help as avoid using some
annoying (and fragile) hacks to locate our actual binary/other resources
Signed-off-by: Seth Flynn <getchoo@tuta.io>
(cherry picked from commit 06e99e2990)
The skin overlay was drawn together with the original skin making it
blend weirdly. By drawing the overlay after the skin this blends with
the body corectly.
Signed-off-by: Trial97 <alexandru.tripon97@gmail.com>
(cherry picked from commit 624d506fac)
