update 1.1.9.4 ( tests )

[egal8888]

Hi Rudi,

please have a look at this old thread:
https://forum.ultravnc.net/viewtopic.ph ... 579#p93579

This was the first time this error occured. And even that you did not answer my post, you fixed that bug in 1.1.7.
So, the best thing might be to have a look at the code changes between 1.1.6 RC and 1.1.7.

This monitor is a single ASUS PB278Q and it has a native resolution of 2560 x 1440 x. No second monitor is in use.

This error occurs with mirror driver enabled and disabled. There's no difference if you use it or not.
And it even does not matter what other graphics related settings you adjust in the ultravnc.ini. You will always get this error.

The only thing to work around this is to replace the ultravnc.exe 1.1.9.4 with the ultravnc.exe of your latest pusblished 1.1.9.3 release and restart the server service. After that you can use any 1.1.x viewer to successfully connect to that machine.
That's why I'm pretty sure there must be a code change between 1.1.9.3 and 1.1.9.4 that is causing this issue. Just ask yourself in which cases you throw that error message mentioned in my two postings.

I also tested FullHD screen resultion at 1920 x 1080. This is working fine using you latest 1.1.9.4 build. Perhaps some variables used might cause a kind of overflow if screen resolutions above FullHD are used. That could explain why ultranvnc thinks that a value for height or width is negative even that I cannot imagine of any data structur that causes overflows at values above e.g. 2048. You are surely using a some kind of integer for storing the values for height and width.

Only you will know

[Rudi De Vos]

That 1920*1080 works fine is a nice indication.
Possible, somewere in the code 2560x1440x32 cause a buffer overflow or is bigger then the var type used.
A buffer overflow behave unexpected, if it's in a memory part that's unused at that point it works als you
can have all kind of errors..

[Rudi De Vos]

I need the logging info from winvnc.exe, can you activate the log.
Possible it indicate the exact function that fail.


width=0 height=0 is return if one of the following functions fail.
GetDC()
CreateCompatibleBitmap()
CreateCompatibleDC()
GetDIBits()
Thoose functions allocate memory to hold a bitmap big enough to to fit the screen display.
If one fails we return a 0 size to stop java.
This part didn't changed for 4 years...

A lot of code was changed between 1193 and 1194
update jpeg library, 50 src files

composed monitors use the same vars as a single, if it was a type error it would also fail for a composed.
I tested it with 3840x1200 and that seems to work on my PC, vars can be excluded.

[egal8888]

Rudi,

sorry for that.
Believe it or not, as of today I cannot reproduce that error.
The past days, when I tried to connect to that machine mentioned, UltraVNC always closed down immediately giving the described error.

Today I replaced the existing ultravnc.exe V1.1.9.3 with the ultravnc.exe V1.1.9.4 as I did several times before, enabled full debugging and restarted the server service.
And now it suddenly works.

I really hate such errors, that occur and go away for no visible reason

Again, sorry to have bothered you. If the error occurs again I will enable debugging immediately and send you the logs.

[Rudi De Vos]

No probs.

Searching MS msdn libs i found that the getdc and other function can fail.

When they fail it's because there are no more OS resources, so it could that after a while
when your OS has been running for a longer period, starting stopping apps..
The resources are no longer there to create the big bitmaps..
no bitmap creating -> vnc close connection.