zturner added inline comments.
================
Comment at: source/Target/Platform.cpp:986-991
+ if (normalized_triple.getVendorName().empty())
+ normalized_triple.setVendor(compatible_triple.getVendor());
+ if (normalized_triple.getOSName().empty())
+ normalized_triple.setOS(compatible_triple.getOS());
+ if (normalized_triple.getEnvironmentName().empty())
+ normalized_triple.setEnvironment(compatible_triple.getEnvironment());
----------------
zturner wrote:
> labath wrote:
> > zturner wrote:
> > > Are these cases even possible? Why would the vendor and os ever be
> > > empty? I thought only the environment could be empty.
> > It is possible, because in some cases, we actually only specify the
> > architecture in calls to this function (see the arch_name comment above).
> > There (I think) its only purpose is to disambiguate which slice in a fat
> > binary are you talking about (and in that case, you don't really need to
> > specify anything other than an architecture).
> >
> > However, in that case, it is true that what we are passing is not really a
> > triple.
> Right, but you are specifically checking against the //normalized// triple.
> I'm pretty sure `Triple::normalize` will never set these to empty strings.
> Maybe I'm wrong though (I'm actually about to test)
Huh. Ok, I guess you're right. The unit tests have this check:
```
EXPECT_EQ("i386", Triple::normalize("i386"));
```
I assumed that would return `i386-unknown-unknown` since they are not
considered "optional" components.
https://reviews.llvm.org/D39387
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
