[PATCH] D39049: [analyzer] Fix wrong calculation of offset in ArrayBoundsV2

[Daniel Marjamäki via Phabricator via cfe-commits]

danielmarjamaki added a comment.

> Could you do a similar analysis that I did above to check why does this not work for the multidimensional case? (I.e.: checking what constraints are generated and what the analyzer does with them.)

the "location.dump()" will just say "x".

the ProgramState is:

Expressions:
 (0x2acd12a78a0,0x2acd1478b80) buf : &buf
 (0x2acd12a78a0,0x2acd1478bf8) buf : &element{buf,0 S64b,int [3]}
 (0x2acd12a78a0,0x2acd1478c10) buf[1] : &element{buf,1 S64b,int [3]}
 (0x2acd12a78a0,0x2acd1478c38) x : &x
 (0x2acd12a78a0,0x2acd1478c88) buf[1] : &element{element{buf,1 S64b,int [3]},0 S64b,int}
Ranges of symbol values:
 reg_$0<int x> : { [4, 10] }

rawOffsetVal => 0

extentBegin => 0

For getSimplifiedOffset() , the offset is not a SymIntExpr it will just return 0.


Repository:
  rL LLVM

https://reviews.llvm.org/D39049