Is there a way to connect to and use X server without calling XOpenDisplay? For example, is it possible to use X rendering capabilities to render to a bitmap? Does all interaction must start with XOpenDisplay?
No, that's not possible. Asking an X server to do things for you necessarily requires talking to an X server.
You can create virtual X servers (connected to no video or input hardware, just rendering into memory) with the Xvfb X server.
Related
I'm dealing with X application behaviors sometimes changing depending on the remote X server (particularly Windows-based ones like eXceed, Xming, FastX, and Cygwin-X). Is there any "simple" way to determine what the remote X server is?
It's possible to use one twain driver to manage concurrent request to two different multifunction printer?
I mean, if I have two MFPs , can I do two scan request in paralel using the same twain driver?
It depends on if your driver supports it.
From the TWAIN Spec page 125:
If an application attempts to connect to a Source that only supports a single connection when the source is already opened, the Source should respond with TWRC_FAILURE and TWCC_MAXCONNECTIONS.
Also from the spec on page 212:
The Source is responsible for managing this, not the Source Manager (the Source Manager does not know in advance how many connections the Source will support).
I tested this with a Fujitsu fi-7260 scanner and got the TWCC_MAXCONNECTIONS error with Twacker:
It could be possible. The reason being TWAIN just sits between the application and the images fed to it.
Imagine a scenario something on the below lines:
1) User clicked on the scan button.
2) You initiate the network layer calls to start the scan job.
3) Now instead to one printer you start scan jobs on two printers from two threads.
4) Let's say each of those threads populate the raw BMP data to a single data structure that is shared.
5) Once both threads are complete iterate over that shared data structure to pass the images to the application via the XFERIMAGE call.
Basic idea is to create an abstraction of two printers behind the scene.
Please let me know if my understanding of your question was not correct or you need other clarification.
If you implement it in the described way, it usually works only with two different MFP's as the majority of TWAIN drivers do not support two different USB devicves at the same time.
I'm writing a bit of desktop software which has two components. Component B queries component A. Creating a web service seems like an ideal way to do IPC in principle. The data model fits, there are ready-made client and server libraries, a well known way to encode and decode parameters etc.
But setting up an HTTP server on a network socket doesn't seem right for a local application. For example what port do I choose? I don't really want people to be able to scan and talk to the app from outside etc.
So I was thinking that I might be able to do HTTP over a domain socket. Does that make any sense? Is there any precedence for it? Is there an equivalent protocol that I could use for IPC which has the same properties as HTTP (requests for specified resources (URIs), encoded parameters, response)?
Looking for C libraries (and possibly Go and ObjC for bonus points).
Binding to the loopback interface only (127.0.0.1) solves your "external visibility" problem, only processes on the local machine will be able to connect.
It does not solve your port allocation problem though, the port number you choose might be taken by the time your app starts. Then your server can't bind and your client connects to the other process bound to your port.
Old, less hip, but CORBA implementations tend to have the problems you have not thought of yet figured out already.
I have a nginx server that has a redis master and two salves of the master. The slaves are read and the master is read and write. Nginx server is fastcgi using spawed python apps and using pyredis.
When is comes for a read from my nginx app, what is best practice for determining which server gets the read among the three? Is it determined in realtime? Do I just do simple random selection using round robin in real time?
Again, I just have on master. Soon I will have two and will use consistent hashing in python using http://pypi.python.org/pypi/hash_ring so select which server gets the keys.
For the interim, is it wise to select which server will get the read using the hash ring even though they should be exact copies?
Thanks,
what you should do is abstract the code that does that so it doesn't change your app logic later when you split the data.
and as for reading - I'd use just the slaves for that. you can use the hashing if you want, provided it doesn't affect your code and is abstracted.
Is there any way to run an NBD (Network Block Device) client and server on the same machine without deadlocking the system?
I am very exhausted looking to find an answer for this. I appreciate if anyone can help.
UPDATE:
I'm writing an NBD server that talks to Google Storage system. I want to mount a file system on the NBD and backup my files. I will be hugely disappointed if I have to end up running the server on another machine. Few ideas I already had seem to lead nowhere:
telling the file system to open the block device using O_DIRECT flag to bypass the linux buffer cache
using a raw device (unfortunately, raw devices are character devices and FSes refuse to use them as underlying device)
Just for the record, having the NBD client and server on the same machine has been possible since 2008.
Use a virtual machine (not a container) - you need two kernels, but you don't need two physical machines.
Since the front page of the Sourceforge project for NBD say that a deadlock will happen "within seconds" in this scenario, I'm guessing the answer is a big "No."
Try to write a more complete question of what actual goal you're trying to accomplish. There's some times that you need to bang away at a little problem, and some times that you need to look at the big picture.