llvm-svn: 72030

and pointer types safely if they only do so when the sizes are the same.
llvm-gcc is such a client.

llvm-svn: 72029

Patch by John McCall.

llvm-svn: 72028

llvm-svn: 72027

llvm-svn: 72026

between integers and pointers when the source type is marked signed,
since inttoptr and ptrtoint always use zero-extension when the destination
is larger than the source.

llvm-svn: 72025

llvm-svn: 72024

llvm-svn: 72023

Some comment cleanup.

llvm-svn: 72022

llvm-svn: 72021

llvm-svn: 72020

 - Chris, please see added FIXMEs.

llvm-svn: 72019

width of a terminal. Don't try to get the width of a terminal if we
don't have this header.

llvm-svn: 72018

llvm-svn: 72016

template, introduce that member function into the template
instantiation stack. Also, add diagnostics showing the member function
within the instantiation stack and clean up the qualified-name
printing so that we get something like:

  note: in instantiation of member function 'Switch1<int, 2, 2>::f'
  requested here

in the template instantiation backtrace.

llvm-svn: 72015

llvm-svn: 72014

llvm-svn: 72013

avoid ambiguity with the word "use" in IVStrideUse.

llvm-svn: 72012

llvm-svn: 72011

bits have been stripped out by instcombine.

llvm-svn: 72010

llvm-svn: 72009

consistency with other routines that use a null Loop* to mean code
not contained by any loop.

llvm-svn: 72008

instead of within their controlling expressions.

llvm-svn: 72007

type mismatches.

llvm-svn: 72006

entering a loop is a non-split critical edge.

llvm-svn: 72004

and similar to ConstantInt's isOne().

llvm-svn: 72003

llvm-svn: 72002

delta tree.

The issue is roughly a conflict in ReplaceText between two kinds of 
uses. One, it should be possible to replace a replacement: for example, the
ObjC rewriter calls ReplaceStmt for an expression, then replaces the resulting
expression with another expression.  Two, it should be possible to 
replace text that already has text inserted before it: for example, the 
HTML rewriter inserts a bunch of tags at the beginning of the line, then 
tries to escape the first character on the line.  This patch 
distinguishes the two cases by storing the deltas separately; 
essentially, replacements and insertions no longer interfere with 
each other.

Another possibility would be to add some sort of flag to ReplaceText, but
this seems a bit more intuitive and flexible.

There are a few downsides to the current solution: one is that there isn't
any way to remove/replace an insertion without touching additional
surrounding text; if such an operation turns out to be useful, an
additional method or flag can be added.  Another is that an insertion 
and replacing a string of length zero are distinct operations; I'm not 
sure how to resolve this, or whether it will be confusing in practice.

This is relatively sensitive code, so please test and tell me if 
anything breaks.

llvm-svn: 72000

It doesn't matter for piped input, but it's annoying when typing at the 
console.

llvm-svn: 71998

llvm-svn: 71997

llvm-svn: 71996

llvm-svn: 71993

ASTConsumers actually do.

llvm-svn: 71992

for PostRAScheduler. 

llvm-svn: 71991

llvm-svn: 71990

llvm-svn: 71989

llvm-svn: 71988

llvm-svn: 71987

when not doing unit-at-a-time.

llvm-svn: 71986

llvm-svn: 71984