I use tigervnc[server] in conjunction with xinetd to fire up a server when someone opens a connection on the specified port. My xinetd.d service looks like the following service myvnc { protocol = tcp socket_type = stream wait = yes user = myuser server = /usr/bin/Xvnc server_args = -geometry 1280x1024 -listen tcp -extension MIT-SHM -inetd -query localhost -once -depth 24 -rfbauth /home/myhome/.vnc/passwd } After upgrading to 1.7.1-r2 this no longer works, the Xvnc process just pegs the CPU at 100% and no connection is established. If I start Xvnc by hand with the command line: Xvnc -geometry 1280x1024 -listen tcp -extension MIT-SHM -query localhost -once -depth 24 -rfbauth /home/myhome/.vnc/passwd :1 It works as one might expect. I am seeing this behaviour on both x86_64 and arm. Reproducible: Always
Is this a regression in comparison to some other version?
Yes, this works fine with previous versions, most recently =tigervnc-1.7.1, so it has to do with the -r2 patchset I'm guessing. Also, the -r2 patches say they are so that you can compile against xorg 1.19, but I had no trouble downgrading back to tigervnc-1.7.1, which was compiled against 1.19. Dunno what that means, but is notable.
Created attachment 470906 [details, diff] Patch from tigervnc github fixing described behaviour I trolled tigervnc's Github and found this commit, which fixed my issue.
(In reply to David Flogeras from comment #3) > Created attachment 470906 [details, diff] [details, diff] > Patch from tigervnc github fixing described behaviour > > I trolled tigervnc's Github and found this commit, which fixed my issue. trawled, not trolled.
I can confirm it, but not only xinetd, also calling via lightdm produce the same bug - 100% cpu load with Xvnc process and nothing. So it is a wider issue. The patch is helps. Thank you David! Yes, it's a regression.
confirmed by Alex. otakuto, what is your opinion about the patch? Is the bug still present in 1.9.0-r1?
Wow, I'd forgotten about this one. I've been using 1.9.0-r1 on arm and amd64 without patches with no trouble for over a year. Probably safe to close this as it was integrated upstream.
(In reply to David Flogeras from comment #7) > Wow, I'd forgotten about this one. I've been using 1.9.0-r1 on arm and > amd64 without patches with no trouble for over a year. Probably safe to > close this as it was integrated upstream. Yep, that commit is in all versions of TigerVNC currently in tree.
