https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115349
Bug ID: 115349
Summary: compiler infers the wrong Accum_Type for a Reducer
expression
Product: gcc
Version: 14.1.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: ada
Assignee: unassigned at gcc dot gnu.org
Reporter: devotus at yahoo dot com
CC: dkm at gcc dot gnu.org
Target Milestone: ---
Created attachment 58349
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58349&action=edit
Minimal (non-)working example of the bug
Full description in the thread that starts on this page:
https://forum.ada-lang.io/t/regression-in-gnat-14/890?u=cantanima
Essentially, a function returns a vector of `Positive`. I call `'Reduce("+",
0)` on the result. This crashes at run time because the compiler infers that
"+" should operate on `Positive`, and 0 is not positive. The compiler does at
least warn me of this.
However, this was not the behavior in 13.x, which inferred that "+" should
operate on either `Natural` or universal integer. Several of us, including
Tucker Taft, agree that the 13.x behavior is consistent with both the wording
and the intent of ARM 4.5.10 (Taft cited (12/5-13/5) specifically). Thus, we
consider it a bug.
I am attaching what I believe is a minimal working example. (Well,
"non"-working, really.)
