On 05/25/2017 02:14 PM, Gregory Szorc wrote:
On Thu, May 25, 2017 at 5:31 AM, Ehsan Akhgari
<ehsan.akhg...@gmail.com <mailto:ehsan.akhg...@gmail.com>> wrote:
What was the motivation behind this change? Was there a complaint
from a significant number of developers about it being difficult
fixing compiler warnings grepping for things like "warning:" or
using ./mach warnings-list? Was the feedback from developers who
have the use case of writing code and fix compiler errors in such
code taken into account?
The motification was that relevant compiler warnings are flying under
the radar and are detected too late in the development process - such
as when pushing to Try. We want to catch problems as early in the
development cycle as possible to reduce the end-to-end times for patch
development. There have been complaints from developers over the years
about this. What specifically motivated me to make this change
recently was finding a pattern of feature requests during bug triage
that all seemed to point to improving visibility of compiler warnings
as a potential solution.
I fully acknowledge that the spew of warnings at the end of the build
is annoying. This was expected to cause temporary pain. The hope was
that this would spur people into fixing the warnings.
FWIW, IMO this is not OK. I don't think there is any consensus that
fixing compiler warnings is suddenly of such a great importance that we
need to interrupt everyone's different workflows without them having
opted into doing the work to fix the warnings. And it may very well
turn out that a project with a different format that isn't structured
around the tools forcing people into fixing things may be more suitable
to efforts like this, as years of work on
https://bugzilla.mozilla.org/show_bug.cgi?id=buildwarning shows.
There has been some movement there. ICU warnings were suppressed, for
example. And, there has been talk in this thread (and presumably
elsewhere) about fixing other large offenders. So on that front, I'd
say this change was successful.
I hope in the future you would consider discussing disruptive changes
like this first instead of announcing them post-facto. Maybe at least we
would have been able to think of a better initial implementation
including not emitting the list for third-party library warnings or
failed builds. Or maybe we should have allowed people to opt in first.
As the bug above shows, there are people who are interested to help in
this area.
But the pace of progress hasn't been terrific and the status quo is
annoying. This thread is also sapping up people's time. I've prepared
a patch to tweak the behavior that I think strikes a compromise. I'll
file a bug to track the review shortly.
Thank you. I think most of the immediate productivity drain issues are
fixed for now.
Cheers,
Ehsan
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform
