[Bug default/29044] New: abidiff reporter inconsistency

["gprocida at google dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29044

            Bug ID: 29044
           Summary: abidiff reporter inconsistency
           Product: libabigail
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: default
          Assignee: dodji at redhat dot com
          Reporter: gprocida at google dot com
                CC: libabigail at sourceware dot org
  Target Milestone: ---

Created attachment 14057
  --> https://sourceware.org/bugzilla/attachment.cgi?id=14057&action=edit
XML inputs and abidiff outputs

Hi Dodji.

While digging into another issue, I noticed an inconsistency between whether
abidiff and abidiff --leaf-changes-only considers there to be a ABI break.

The attachment contains two Android kernel XML files. An independent check
shows that the only differences between them are whether certain types are only
declared or fully defined.

1. abidiff reports that ir_raw_handler_register's type has changed.

2. abidiff --leaf-changes-only does not.

3. abidiff --harmless results in reports of lots of cases where only-declared
vs fully-defined is considered to be a harmless change.

4. abidiff --leaf-changes-only --harmless results in a single report that
struct drm_crtc_helper_funcs has been changed, but the report doesn't reach a
root cause.

In general, the inconsistency is a bit surprising, however, I think the best
places to look at in terms of bug hunting are 1 and 4 as, ignoring
only-declared vs fully-defined issues, nothing has changed between the two
ABIs.

Regards,
Giuliano.

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