[Lldb-commits] Exceptions with builder lldb-x86_64-linux

[Greg Clayton]

On May 7, 2013, at 1:22 PM, "Malea, Daniel" <daniel.malea at intel.com> wrote:

> The exceptions are just another way of saying the tests "failed". It's not great but it doesn't seem to interfere with error reporting or detection. Once the failures are resolved, the buldbot should be green. We should, of course, look at what is causing the buildbot to go into "exception" state rather than "failed" but the point is that we still have failures that are higher priority.
> 
> The watchpoint errors were recently introduced a few commits ago; Matt is looking at those.
> 
> But there is also the Test13338477.py which has been failing for a long time when built with GCC (but not clang). I looked at it with Matt just now, and the failure is unrelated the original "de-sugaring" fix that was implemented in the regressing commit (r180938). Instead, it seems that the illegal opcode GCC 4.6 inserts when compiling a "__builtin_trap()" intrinsic messes with LLDB's ability to unwind the stack, thereby causing problems evaluating the local variable foo. We should split up that test case so we can file a bugzilla for the new bug.
> 
> That said, that test also relies on a bug (or feature?) allows LLDB to print variables that are not in the selected frame. Can anyone confirm if this is intended behaviour? Seems like it would be cause for much confusion. What I mean is if you have:
> 
> Void bar() {
>  // break here
> }
> 
> Int main() {
>  Float foo 1.f;
>  Bar();
>  Return 0;
> }
> 
> If LLDB is stopped at "break here" (I.e. The frame containing 'bar' is selected) is the user supposed to be able to run "print foo" and expect a valid result? Because that's what's happening right now, at least in Test13338477.py.

This shouldn't be happening. "expression" will look for the locals/args in the current frame, then for file statics or globals from the current frame's compile unit, then global variables from any files in all other modules.

I would look at which "foo" is is finding as it may have found a function foo, or incorrectly found "foo" from somewhere else. The first thing to do is to find out which foo it found, then we can track down why it found it.


> 
> 
> Cheers,
> Dan
> 
> 
> From: Enrico Granata <egranata at apple.com<mailto:egranata at apple.com>>
> Date: Tuesday, 7 May, 2013 4:04 PM
> To: Ashok Thirumurthi <ashok.thirumurthi at intel.com<mailto:ashok.thirumurthi at intel.com>>
> Cc: "lldb-commits at cs.uiuc.edu<mailto:lldb-commits at cs.uiuc.edu>" <lldb-commits at cs.uiuc.edu<mailto:lldb-commits at cs.uiuc.edu>>
> Subject: Re: [Lldb-commits] Exceptions with builder lldb-x86_64-linux
> 
> I might be wrong, but at a glance it looks like all these errors are related to watchpoints.
> It might well be that something on the hardware of the buildbot has watchpoint issues - not my expertise area and hard to be specific. Just a pointer.
> 
> Enrico Granata
> ✉ egranata@.com
> ✆ 27683
> 
> On May 7, 2013, at 12:57 PM, "Thirumurthi, Ashok" <ashok.thirumurthi at intel.com<mailto:ashok.thirumurthi at intel.com>> wrote:
> 
> Any thoughts on what's causing the exceptions with the gcc lldb buildbot at http://lab.llvm.org:8011/builders/lldb-x86_64-linux?  We haven't been able to reproduce these with Ubuntu 12.04 (python 2.7.3) or 12.10.  Which version of python 2.7 is installed?
> http://lab.llvm.org:8011/builders/lldb-x86_64-linux/builds/3961/steps/test%20lldb/logs/err.text
> 
> - Ashok
> 
> -----Original Message-----
> From: lldb-commits-bounces at cs.uiuc.edu<mailto:lldb-commits-bounces at cs.uiuc.edu> [mailto:lldb-commits-bounces at cs.uiuc.edu] On Behalf Of Duncan Sands
> Sent: Thursday, November 08, 2012 3:07 AM
> To: Galina Kistanova
> Cc: lldb-commits at cs.uiuc.edu<mailto:lldb-commits at cs.uiuc.edu>; llvm-commits at cs.uiuc.edu<mailto:llvm-commits at cs.uiuc.edu>; llvm cfe
> Subject: Re: [Lldb-commits] Removing builder lldb-x86_64-linux?
> 
> Hi Galina,
> 
> I am going to remove builder lldb-x86_64-linux from llvm buildmaster,
> Seems no one is watching the builder, it fails for a very long time.
> 
> it's a pity that none of the LLDB people seem to care, but since they don't please feel free to remove it.
> 
> Ciao, Duncan.
> 
> If there are any objections, please speak up!
> 
> Here is the builder:
> http://lab.llvm.org:8011/builders/lldb-x86_64-linux
> 
> Thanks
> 
> Galina
> 
> 
> _______________________________________________
> lldb-commits mailing list
> lldb-commits at cs.uiuc.edu<mailto:lldb-commits at cs.uiuc.edu>
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits
> 
> _______________________________________________
> lldb-commits mailing list
> lldb-commits at cs.uiuc.edu<mailto:lldb-commits at cs.uiuc.edu>
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits
> 
> 
> _______________________________________________
> lldb-commits mailing list
> lldb-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits