On Thu, Oct 06, 2016 at 05:58:10PM +0300, Ponomarenko Andrey wrote:
06.10.2016, 08:23, "Pierre-Yves Chibon":
> On Wed, Oct 05, 2016 at 06:36:16PM +0300, Ponomarenko Andrey wrote:
>> The tool is based on different software stack for analysis of backward
>> compatibility developed since 2009:
https://github.com/lvc (ABI Compliance
Checker, ABI Dumper, etc.)
>>
>> RedHat created an alternative libabigail tool in 2013. Implementation and
reports are completely different. But anyway, two is better than one. Now we can verify
reports of both tools by each other.
>
> I'm confused what Red Hat as to do in there. As far as I know, it's a person
not
> a company that runs the development or libabigail and I very much doubt that
> this person was tasked to do that by some higher power.
>
> That being said, did you look at it? Did you make some comparison on how it
> performs compared to this stack you mention?
> Are there times where one finds something that the other don't, vice-versa?
> Can they be ranked or are they too different to be compared?
>
> Thanks,
> Pierre
After a closer look at the source code, reports and docs of abipkgdiff / libabigail tools
I can list some pros and cons of
https://github.com/lvc/pkg-abidiff /
abi-compliance-checker:
PROS
- separated analysis of both backward binary compatibility and backward source
compatibility
- assigning severity levels to ABI changes
- explaining effects of ABI changes
- checks for more compatibility rules
- less false positives
Can you quantify these last two? Are we speaking about 1 or 2 more checks or
much more?
Are we speaking about 1 or 2 more false positives or much more? Over how many
packages?
Also, while I'm not involved in either projects, did you report this to the
libabigail maintainers? I'm sure he would welcome bug reports and test-cases for
their project.
Maybe there is a way to collaborate in such way that both projects improve :)
Thanks,
Pierre