On Mon, Mar 19, 2012 at 1:35 PM, Lee Essen <lee.es...@nowonline.co.uk> wrote: > On 19 Mar 2012, at 12:32, Andreas Färber wrote: > >> Am 19.03.2012 13:05, schrieb Lee Essen: >>> On 19 Mar 2012, at 11:59, Stefan Hajnoczi wrote: >>> >>>> I'm going to spend some time today reviewing recent tracing patches. >>>> I'd prefer to move to a Python version of tracetool rather than worry >>>> about the shell quirks across all host platforms. >>>> >>>> If the Python rewrite cannot be merged for 1.1 then it makes sense to >>>> go with shell portability fix. >>> Hi Stefan, >>> >>> While you are looking at this, there are a couple of other related issues >>> worth having in the back of your mind: >>> >>> 1. "self" is a reserved word in Solaris/Illumos trace, and it's used in a >>> few trace calls. >>> 2. "bool" isn't recognised by default, again used in a couple of traces ... >>> could be fixed by typedef, but switching to int is probably better (imho) >>> 3. Some work is needed on the linking stage when using Solaris/Illumos >>> dtrace. You need to provide all the objs to the dtrace -G call in order to >>> get an object generated that includes all of the relevant symbols in it. >>> Happy to provide more detail if needed. >> "Reviewing recent tracing patches" does not imply writing patches to fix >> everyone's issues, you'll likely need to send patches for those issues >> yourself at some point (or me if I find time, or someone else). >> >> Stopping to use bool throughout QEMU is not an option IMO. If it's >> limited to the DTrace backend code and doesn't negatively affect >> SystemTap then that may be an option. > > > In my original (way-too-long) patch for Illumos I included fixes for both of > the above, the approach I took was to map "bool" to "int" and "self" to > "_self" in the dtrace part of tracetool. So it should have had no impact on > any non-dtrace stuff. > > If that seems like a sensible approach I will look again, and at the linking > issues.
That sounds like it can work. Stefan
