[Ada] Clean up in handling of Bound expressions

[Arnaud Charlet <charlet at adacore dot com>]

Tested on x86-linux, committed on mainline.

The call to resolve the Bound expression in function One_Bound was
premature, and resulted in bounds expressions with type universal_integer
rather than the type of the range in some cases. This was happening in
particular when literal bounds should resolve to Integer. The resolution
of the bounds needs to be delayed until the call to Make_Index, which
accounts for the case where literal range bounds should resolve to Integer.
This turned out not to cause a problem for the gcc-based GNAT compiler,
but was problematic for the non-gcc-based AAMP back end, which expects
integer literals to have their type set properly.

2004-12-07  Gary Dismukes  <dismukes@adacore.com>

	* sem_ch5.adb (One_Bound): Remove call to Resolve, as the bound needs
	to be resolved later as part of Analyze_Iteration_Scheme's call to
	Make_Index.

Attachment: difs.21
Description: Text document