<div dir="ltr">Ok, thanks for taking a look. I'll have a look again and see if I can expose something.</div><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Dec 12, 2014 at 6:56 PM, Duncan P. N. Exon Smith <span dir="ltr"><<a href="mailto:dexonsmith@apple.com" target="_blank">dexonsmith@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Oversight. TBH, I didn't look at your patch again; I just saw the PR,<br>
remembered you'd fixed the problem before, and looked up the thread in<br>
order to credit you (and to find the right paint colour).<br>
<br>
Just now, I tried to expose a few other problems with testcases, and they<br>
all seem to be invalid (`llvm-as` doesn't accept strange characters in<br>
metadata names) or already work (names of globals, structs, and locals).<br>
<br>
Go ahead and commit (or file a PR) for anything you can make a testcase<br>
for though!<br>
<div class="HOEnZb"><div class="h5"><br>
> On 2014 Dec 12, at 15:08, Keno Fischer <<a href="mailto:kfischer@college.harvard.edu">kfischer@college.harvard.edu</a>> wrote:<br>
><br>
> Hi Duncan,<br>
> looking at your commit, were the other three instances that were in my patch of this pattern not affected, or was that an oversight?<br>
><br>
> Thanks,<br>
> Keno<br>
><br>
> On Thu, Dec 11, 2014 at 6:53 PM, Duncan P. N. Exon Smith <<a href="mailto:dexonsmith@apple.com">dexonsmith@apple.com</a>> wrote:<br>
> No problem!<br>
><br>
> > On 2014 Dec 11, at 15:52, Keno Fischer <<a href="mailto:kfischer@college.harvard.edu">kfischer@college.harvard.edu</a>> wrote:<br>
> ><br>
> > Sorry about that. I haven't found the time to get back to this (super busy this last month). I was planning to look at this again next week, but thanks for committing it :)!<br>
> ><br>
> > On Thu, Dec 11, 2014 at 6:38 PM, Duncan P. N. Exon Smith <<a href="mailto:dexonsmith@apple.com">dexonsmith@apple.com</a>> wrote:<br>
> ><br>
> > > On 2014 Nov 9, at 13:17, Keno Fischer <<a href="mailto:kfischer@college.harvard.edu">kfischer@college.harvard.edu</a>> wrote:<br>
> > ><br>
> > > Previously when a metadata string contained unicode characters,<br>
> > > it would be incorrectly placed in the Record array because chars<br>
> > > are signed by default and hence characters with the high bit set<br>
> > > would get sign extended, but the bitcode writer was attempting<br>
> > > to write the lowest 8 bit of the now sign-extended value. This<br>
> > > caused an assertion failure later on. The fix is just to cast<br>
> > > the pointer to uint8_t* first to prevent sign extension.<br>
> > > This came up in the context for metadata strings, but I did a<br>
> > > quick pass and changed the other instances of this pattern in<br>
> > > the file as well.<br>
> > ><br>
> > > <a href="http://reviews.llvm.org/D6184" target="_blank">http://reviews.llvm.org/D6184</a><br>
> > ><br>
> > > Files:<br>
> > > lib/Bitcode/Writer/BitcodeWriter.cpp<br>
> > > test/Bitcode/unicode.ll<br>
> > > <D6184.15959.patch>_______________________________________________<br>
> > > llvm-commits mailing list<br>
> > > <a href="mailto:llvm-commits@cs.uiuc.edu">llvm-commits@cs.uiuc.edu</a><br>
> > > <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a><br>
> ><br>
> > This review seemed to get stalled. I just committed a fix based on Keno's<br>
> > patch in r224077 (which I think matches what David and my bike-shedding<br>
> > landed on), fixing PR21882.<br>
> ><br>
><br>
><br>
<br>
</div></div></blockquote></div><br></div>