Firm Bugtracker - libFirm
Viewing Issue Advanced Details
100 analyses minor always 2013-04-17 13:47 2016-01-03 03:30
haass  
Matze  
normal  
closed 1.21.0  
fixed  
none    
none  
0000100: heights are miscalculated
heights are miscalculated for nodes with successors > 1
in ir/anan/heights.c:148
 ir->height is increase for every available successor

suggested fix attached
? file icon patch_heights_bug.patch [^] (1,157 bytes) 2013-04-17 13:47
Issue History
2013-04-17 13:47 haass New Issue
2013-04-17 13:47 haass File Added: patch_heights_bug.patch
2013-04-17 15:52 Matze Note Added: 0000152
2013-04-17 15:52 Matze Assigned To => Matze
2013-04-17 15:52 Matze Severity major => minor
2013-04-17 15:52 Matze Status new => resolved
2013-04-17 15:52 Matze Resolution open => fixed
2013-04-17 15:52 Matze version development => 1.21.0
2016-01-03 03:30 Matze Status resolved => closed

Notes
(0000152)
Matze   
2013-04-17 15:52   
Thanks for taking the time to report this bug.
You are right, the algorithm for computing the heights was messed up and produced a strave overaproximation. As we only use heights for speeding up reachability queries in heights_reachable_in_block() nobody noticed this yet.
I fixed the issue in a similar way as your proposed patch in ec9575be3b0f12ee2c65ea874f998ffec4a283d3.