Maze for Linux

Implemented support for epoll_wait. Please try the new release 1.0-beta-2012.02.12, and let me know how it works.

Going to look into remaining problems shortly.

Hello aberardino,

Thank you very much for the detailed report! I will fix all 3 bugs, but let me start with the epoll_wait, because this is the one that prevents you from running maze. It may take me about a week to implement.

Would it be possible for you to send me your binary so I could run the test case? My email is roni at kloobok dot com.

What kind of file descriptors do yo use with epoll - pipes, sockets?

Thanks,

Roni.

The feature is implemented. Please download the latest maze 1.0-beta-2011.11.09.

Thank you very much for providing the binary and all details of the case. Those were of great help in reproducing and fixing the problem!

Hi khegeman,

Thank you for your post. The assertion failure is due to an unsupported debug rule used by the compiler.

I am out of the country now. Unfortunately, I do not have the resources to fix the problem here. I'll fix it first thing when I am back on November 7. Meanwhile, would it be possible for you to send me your binary? My email address is roni at kloobok dot com

Roni.

I implemented support for poll system call; also for non-blocking epoll_wait. The new release 1.0-beta-2011.10.20 is ready for download.

If you try to run the test case again and let us know the result, it will be of great help. Thank you in advance!

The support for poll can be added by Monday.

epoll_create and epoll_ctl are fine; so is non-blocking epoll_wait (i.e. with timeout=0).

How soon you will need epoll_wait? It might take a bit longer. What kind of file descriptors do yo use with epoll - pipes, sockets? Do you use the EPOLLET (edge-triggered) flag when calling epoll_ctl?

Thank you for the info and for the prompt response! I'll look into this right away, and then will give you an estimate of how long it will take us to implement.

Hi hfinucane,

Thanks a lot for reporting the problem. Indeed there are several unsupported system calls, but we can implement the support, no problem. As of today the unsupported calls are:

SYS_poll
SYS_epoll_wait
SYS_ppoll
SYS_epoll_pwait
SYS_epoll_wait_old
SYS_mq_timedsend
SYS_mq_timedreceive
SYS_rt_sigtimedwait
SYS_vfork
SYS_execve


Do you need support for the entire "poll" family, or you could name a specific call(s)?

Hi Nithesh,

Thank you for your patience. The issues are fixed.

Please download the entry marked with from the Downloads area.

It is a special build for you. Please try it and let me know how it works!

I built your system equivalent, reproduced the bug, and fixed it.

Previously 32 bit maze has been tested only on x86_64 machines. On those the 32 bit system calls used "sysenter" instruction. On i686 it was "int $0x80" though. That was why the assertion failed: it did not recognize the correct instruction code in your process' address space.

However I found several problems which you may encounter when running maze on i686 machine. Please bear with me for another few days. I will fix the issues and then update the 32 bit maze.

Thanks! We will build the system and try to reproduce the problem. This may take several days. I'll keep you posted.

Thank you, Nithesh.

Could you also tell me if the OS is 32 bit, and if the Red Hat Enterprise release is 5 or 5.<something>? The two shell commands above would have given you the complete info...

Hi Nithesh,

Thanks a lot for reporting the bug!

From the info I can see that your application is 32 bit, and, that, most likely, your kernel is newer than 2.6.27. The problem has been overlooked mostly because 90% of maze users run 64 bit applications, that no longer use the "sysenter" machine instruction.

Please post the output of these two shell commands:

uname -a

cat /etc/*-release

Having this information about you system will enable us to to fix the bug.

Cheers,

Roni.

It is already available. I sent a mass mail notification yesterday. Did it get through? Please let me know... Anyway, please download the latest version 1.0-beta-2011.04.18.

Have fun!

Roni.

Hi Clay,

Please ignore the message for now. Your binary is fine.

The reason maze gives this message is because you use a new version of g++. There was a small change in the debugging information layout created by the new g++. I did not adjust maze accordingly yet.

Thank you for pointing this out, I will update maze today.

Best,

Roni.

Hello and welcome to the forum!

If you tend to be an early adopter of new software, do try maze, and tell us what you like, and, especially, what you do not like about the product.

Please read our short User Manual to learn how maze works and see if it can help you with your work.

Thank you!

Roni Simonian