Red Hat Bugzilla – Bug 1308810
Weird colours of remote desktop via freerdp on ppc64 and s390x
Last modified: 2017-10-05 07:57:02 EDT
+++ This bug was initially created as a clone of Bug #1306576 +++
Description of problem:
Weird colours of remote desktop via freerdp on ppc64 and s390x architectures.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. xfreerdp $WindowsIP
colours of remote desktop are deformed, maybe due to different endianness
colours of remote desktop are normal
This is arch specific bug, appears on ppc64 and s390x; x86_64 and i386 are fine.
--- Additional comment from RHEL Product and Program Management on 2016-02-11 11:41:21 CET ---
Since this bug report was entered in bugzilla, the release flag has been
set to ? to ensure that it is properly evaluated for this release.
--- Additional comment from Tomas Hudziec on 2016-02-11 11:41 CET ---
--- Additional comment from Tomas Pelka on 2016-02-12 08:00:54 CET ---
--- Additional comment from Oded Gabbay on 2016-02-12 08:13:45 CET ---
I'm missing some info about the setup.
What is the client and what is the remote machine ? Are they both ppc64 ?
Please describe the steps you did both on the client and the remote machine (steps to setup the remote display server)
--- Additional comment from Tomas Pelka on 2016-02-12 08:36:08 CET ---
Tomas was connected from x86_64 to ppc4/s390x via VNC and in VNC session freerdp was executed.
--- Additional comment from Ondrej Holy on 2016-02-12 10:28:33 CET ---
I see weird colors with 8/16/24 color depth also. 32bit color depth (-a 32) seems to work correctly...
Yes probably endianess issue.
I used ssh -X to connect from x86_64 to ppc64 machine and see this issue when accessing Windows 7 machine over rdp.
--- Additional comment from Ondrej Holy on 2016-02-12 16:08:47 CET ---
I hoped it is fixed upstream, but it seems it is even worse. I tried several newer versions and the result is almost same, or it exits with decryption error, or it crashes with segfaults instead.
Several upstream reports exists in regards to color/endianness problems. Probably the most relevant one is following:
I proposed several upstream patches though they are not merged yet and I cooperate with upstream in order to fix FreeRDP under big endian. We decided to defer this for 7.4 due to its complexity. It may be fixed also as a part of the rebase, Bug 1291254.
Bug occurs also with package version freerdp-1.0.2-10.el7 on architectures ppc64 and s390x.
Unfortunately, the color-related fixes were not merged. I discussed this with upstream several times, but my suggestions were not taken into account. I have to propose new patches on top of recent color-overhaul. But I am afraid that fixing this bug would require rebase anyway.
Big endian fixes for colors have been merged upstream recently:
Although the handling in master is totally different, the approach which was finally used upstream should not be so hard to use in downstream also...
I tested the fix with Windows 7 server with all combinations of -a 8/16/24/32, --gdi sw/hw, on X server with depth 16/24/32 and LE/BE systems. The colors on BE systems should be the same as on LE with that fix. It doesn't mean that the colors are 100% correct. There are some problems with -a 8 (Bug 1357558) and also with X server depth 32 (https://github.com/FreeRDP/FreeRDP/issues/4134).