[Bug default/26807] New: Issues with enum-underlying types

[Bug default/26807] New: Issues with enum-underlying types

[gprocida+abigail at google dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=26807

            Bug ID: 26807
           Summary: Issues with enum-underlying types
           Product: libabigail
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: default
          Assignee: dodji at redhat dot com
          Reporter: gprocida+abigail at google dot com
                CC: libabigail at sourceware dot org
  Target Milestone: ---

Hi.

I've posted an RFC patch which illustrates a few issues with libabigail's
treatment of enum-underlying types.

https://sourceware.org/pipermail/libabigail/2020q4/002788.html

Repeating them here:

- signedness and changes thereof (as seen in DWARF) are not modelled
- size and alignment changes are reported twice
- size changes are not considered ABI incompatibilities
- *all* enum-underlying types are currently considered anonymous
  and have the same name, even when they differ in size, alignment or
  signedness                                                                    

The last item is closely related to the others but is troublesome in a
different way. It can cause non-deterministic ordering of such types in XML
output (in the presence of concurrent heap-allocation by other threads). In
particular, it causes test flakiness.

I've not attempted to find a way that signedness changes can cause API
surprises. However, there are mentions of undefined behaviour that may occur
when casting in https://en.cppreference.com/w/cpp/language/enum.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/26807] Issues with enum-underlying types

[gprocida+abigail at google dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=26807

--- Comment #1 from Giuliano Procida <gprocida+abigail at google dot com> ---
Since reporting this, the artificial names of underlying types now include the
size (though forward declared enums end up with types named ...-0).

-- 
You are receiving this mail because:
You are on the CC list for the bug.