diff options
| -rw-r--r-- | gcc/ada/sem_attr.adb | 20 |
1 files changed, 19 insertions, 1 deletions
diff --git a/gcc/ada/sem_attr.adb b/gcc/ada/sem_attr.adb index 79a6f020feb..768cdf63ee4 100644 --- a/gcc/ada/sem_attr.adb +++ b/gcc/ada/sem_attr.adb @@ -8089,8 +8089,26 @@ package body Sem_Attr is -- Arr (X .. Y)'address is identical to Arr (X)'address, -- even if the array is packed and the slice itself is not -- addressable. Transform the prefix into an indexed component. + -- Note that the transformation is safe only if we know that - -- the slice is non-null. + -- the slice is non-null. That is because a null slice can have + -- an out of bounds index value. + + -- Right now, gigi blows up if given 'Address on a slice, and + -- this covers up that bug in one case, but the bug is likely + -- still there in the cases not transformed by this code ??? + + -- It's not clear what 'Address *should* return for a null + -- slice with out of bounds indexes, this might be worth an ARG + -- discussion ??? + + -- One approach would be to do a length check unconditionally, + -- and then do the transformation below unconditionally, but + -- analyze with checks off, avoiding the problem of the out of + -- bounds index. This approach would interpret the address of + -- an out of bounds null slice as being the address where the + -- array element would be if there was one, which is probably + -- as reasonable an interpretation as any ??? declare Loc : constant Source_Ptr := Sloc (P); |
