Messages in this thread | | | Date | Tue, 7 Apr 2009 14:14:57 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH] intel-iommu: fix build with CONFIG_BRANCH_TRACER=y |
| |
* David Woodhouse <dwmw2@infradead.org> wrote:
> On Tue, 2009-04-07 at 11:02 +0200, Ingo Molnar wrote: > > for_each_active_iommu() and for_each_iommu() uses some tricky > > C that is weird and borderline valid but does not allow the > > macro evaluation trick used by the branch tracer/profiler: > > > > drivers/pci/intr_remapping.c:680:28: error: macro "if" passed 2 > > arguments, but takes just 1 > > > > Switch it to a braced group statement. > > It doesn't even need the braces -- just putting simple parens > around the comma-expression would have sufficed (that was Andrew's > attempt). > > But it's a workaround, not a fix -- if we're going to #define if() > then we should damn well make it transparent, and not have to work > around breakage in arbitrary places.
Well, i consider it a feature that it flags weird if (x, y) constructs: and yes, these iterators you introduced, while they are legit C, definitely count as 'weird'. If regular code was doing it, not a loop abstraction, i'd call it non-obvious and borderline broken straight away.
We should _never ever_ put comma statements into if () constructs without a _really_ good reason - and if yes, we can flag that we know what we are doing, via extra parentheses.
> Hence the patch I sent to Linus last night, which fixes it in the > _right_ place: > http://marc.info/?l=linux-kernel&m=123907505308502&w=2
I dont think that's the right fix for the reason above - all the places where the branch-profiler 'broke' the build before were genuinely weird pieces of code that needed fixing.
Ingo
| |