From: Jeremy Carroll (jjc_at_hplb.hpl.hp.com)
Date: Mon May 24 1999 - 09:04:34 BST
James Rouleau/CAM/Lotus wrote: > > Are there any plans for UPS to support threads for those platforms which > have them? I looked at this problem many years ago (about 6). I got alpha code working for Isis threads on HP-UX. (No you've never heard of them!). I think I was working on top of a bastardised UPS 2.45. I think there were a small number of fundamental issues: A: an abstract API into the threads system. [To support a new threads system or a new platform this is what would need re-implementing]. This is the most important piece of hindsight. B: do you wish to leave other threads running while debugging a particular thread, and how to do this in your particular thread package. [This was a must for the particular program I wanted to debug. Everything collapsed in a heap if certain critical threads did not progress in real-time]. This is an advanced feature, too advanced for a first cut, but the API [A] needs to be extensible to cover what's needed here. C: thread-specific breakpoints. [Again not necessary in a first cut, but the GUI design could consider this issue. A way, would be too provide a function in the c-interpteter that gave you the thread ID in an UPS meaningful way.] D: I had to link the program to be debugged with another c file that defined some of the functions needed by the thread debugging API [A]. This was clunky. I am not sure what features thread-aware debuggers give. I've only used a Java one, which doesn't have either B or C. If anyone wants to look at my source, let me know by e-mail, and I'll dig it out. But do remember the bits are old and tired, and were never finished. Jeremy Carroll --- Disclaimer: a private communication of Jeremy Carroll, not an official communication of Hewlett-Packard.
This archive was generated by hypermail 2.1.4 : Wed Feb 13 2002 - 21:51:32 GMT