qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [DISCUSSION] GCOV support


From: Peter Maydell
Subject: Re: [DISCUSSION] GCOV support
Date: Thu, 18 Jun 2020 21:50:30 +0100

On Thu, 18 Jun 2020 at 20:41, Aleksandar Markovic
<aleksandar.qemu.devel@gmail.com> wrote:
> четвртак, 18. јун 2020., Aleksandar Markovic 
> <aleksandar.qemu.devel@gmail.com> је написао/ла:
>> You may recall that I signalled on couple of occasions that there are some 
>> problems related to gcov builds in out-of-tree builds.
>>
>> It turned out that those problems manifest on some opder Linux distribution, 
>> and are always related to the gcovr being older than 4.1. For older gcovr, 
>> the tool simply doesn't connect properly executable and its source files, 
>> and no coverage report is generated (or perhaps only some small portions, 
>> but, on any case, gcov builds are virtually unusable).

Ah. Thanks for tracking this down.

>> I propose that we don't bother supporting systems with gcovr older than 4.1. 
>> We could check version of gcovr in confugure, and refuse gcov builds if that 
>> version is older than 4.1.

Seems potentially reasonable. We don't actually check for gcovr at all in
configure right now...

It looks like we only use gcovr in creating the coverage-report.html --
I guess in theory if you wanted to use gcov directly and didn't care
about the coverage report you could still do that without a new gcovr
(ie if you were just using the facilities we provided before commit
fe8bf5f62972 in 2018). But then we'd have to make the handling of the
coverage report conditional on "do we have gcovr". I don't use gcov
so I don't have any idea whether "use gcov data but don't bother with
the gcovr coverage report" is a useful thing for anybody to be doing
or if it's silly and not worth the effort to try to support.

https://repology.org/project/gcovr/versions has the distro
coverage of gcovr versions. I note that Ubuntu Bionic only
has 3.4 still. But for a developer-use-only tool we can
be a bit less strict about our supported-distros policy I think.

Side note: the coverage-report.html targets probably ought
to be only allowed if we have gcov/gcovr enabled.

thanks
-- PMM



reply via email to

[Prev in Thread] Current Thread [Next in Thread]