On Mon, Jul 16, 2012 at 08:27:50PM +0000, Adam Conrad wrote:
> PS: (FWIW, the original patch was mine, not Luca's, it seems to have
> gotten mis/re-attributed along the way...)
Disregard this bit, BTW. Luca told me on IRC that we arrived at our
shockingly similar workarounds independently.
Anyhow. More from said IRC discussion, while I was brain-dumping
at Luca:
< infinity> DktrKranz: If you feel like debugging further, I *suspect*
the error either lies in the bundled tre (in src/extra/tre),
or in the code in R itself that's calling it.
< infinity> DktrKranz: 20 to 1 odds say it's an alignment issue or
something.
< infinity> DktrKranz: And, for added fun, the bundled tre is a forked
version from upstream tre. So, it could just be that it was
subtly broken in the fork.
< infinity> DktrKranz: It still *could* be eglibc's fault, as we're also
calling into eglibc to mangle wchars and such, but if wchar
was fundamentally broken in eglibc on ARM, I think I'd have
seen a lot more fallout.
< infinity> DktrKranz: So, it's probably either R mis-mangling types or
doing dodgy pointer math, or tre doing similar.
--
To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
