How to capture diag info from a running application?
10/01/2018 14:54
Folks, In my old mainframe days (IBM mainframe that is) we sysprogs used to be able to set what was called a SLIP trap, which would let you say if an instruction is fetched at virt address x, or storage is altered at virtual address x or a branch is taken at virtual address x then do one of more various things like take a dump, trace some data to the system trace facility, stop the whole system, etc.

It there a similar tool, facility, scheme on Solaris? Ideally I'd like to be able to say if my application gets to instruction addr x then do y. I'm looking to do this at a customer site to capture data about an incident that we are unable to repro in house.

I suppose I could just have the customer run the application under gdb (or dbx) and set a breakpoint or whatever via that. What is the cost of running under a debugger if an event hardly ecver happens? Chris

Source is Usenet: comp.unix.solaris
Sign in to add a comment

Answer score: 5
10/01/2018 14:54 - In article <qIGdndl0g4KV3XLdRVn-sw@comcast.com>, Chris Markle <cmarkle@sendmail.com> wrote: I've always called this debugging or profiling. You put conditional code into source that you compile in to output error messages, tell you where the code is spending cycles, etc. Then you run it through the test suite and see if it triggers the bug or under load to collect the profile info to optimize. You deliver code without this feature enabled.

I suppose you could write kernel module that responds to an unimplemented trap and put that trap throughout your code. But Solaris doesn't have such a feature that I'm aware of. Would you customer run a modified kernel and collect debug information for you? Certainly not in production, I hope.

This approach is pretty much a mainframe virtual machine model. Don't know if you can do it on Linux on an IBM system under VM. Might be interesting.

-- DeeDee, don't press that button! DeeDee! NO! Dee...


Source is Usenet: comp.unix.solaris
Sign in to add a comment

Answer score: 5
10/01/2018 14:54 - Michael Vilain <vilain@spamcop.net> wrote: You should probably investigate the pid provider in Dtrace in Solaris Express (See Bryan Cantrill's blog page for links - http://blogs.sun.com/bmc ).

-- Alan Hargreaves - http://blogs.sun.com/tpenta Senior Technical Support Specialist/VOSJEC Engineer Product Technical Support (APAC) Sun Microsystems

Source is Usenet: comp.unix.solaris
Sign in to add a comment

eDiscover
Helpforce eDiscover provides technical articles updated each dayHelpforce eDiscover RSS feed contains the latest technical articles in RSS
Click the logo to go back to the main page
Search eDiscover
  
Categories

Click an icon to go to that category

Helpforce eDiscover contains articles about Microsoft Windows Helpforce eDiscover contains articles about Apple products and MacOS Helpforce eDiscover contains articles about Linux and POSIX operating systems Helpforce eDiscover contains articles about Helpforce Helpforce has a large variety of technical information and articles for you to read Helpforce eDiscover contains articles about databases, MYSQL, SQL Server Oracle Helpforce eDiscover contains articles about Java, JVM and the JRE Helpforce eDiscover contains articles about the QNX operating system Helpforce eDiscover contains articles about Oracle Solaris and Open Solaris Helpforce eDiscover contains articles about RISC OS, Acorn and the BBC Micro Helpforce eDiscover contains articles about Amiga and AmigaOS

Type your comment into the box below