Formatting on Windows

[Thomas Stüfe]

On Sat, Jul 29, 2023 at 2:45 PM David Holmes <david.holmes at oracle.com>
wrote:

> Hi Julian,
>
> On 29/07/2023 3:06 pm, Julian Waters wrote:
> > As discussed in https://github.com/openjdk/jdk/pull/15063
> > <https://github.com/openjdk/jdk/pull/15063>, how many of these are
> > formatting errors worth fixing, and how should they be fixed if they
> > have to be? gcc has flagged all of them as formatting errors, and while
> > I could simply silence the checker in compilerWarnings_gcc.hpp on
> > Windows, I'm unsure if that's the right course of action to take
>
> Where was it determined that building for Windows using gcc was a goal
> of OpenJDK? This is (as Thomas tried to point out) effectively another
> port, so who is testing and supporting it? I can understand it is an
> interesting side project for you, but we all end up with the burden of
> having to deal with this - first by reviewing these PR's and then by
> dealing with future breakage. AFAIK no one delivering OpenJDK is using
> gcc for a Windows build. You can't expect everyone to check whether a
> change that is fine with VS on Windows is also fine with gcc, so this
> will be forever breaking with ensuing PRs (that others have to expend
> effort reviewing) to fix it again.
>
> Sorry to dampen your enthusiasm on contributing here but there has to be
> a strong reason to make these kinds of changes. Who wants/needs this?
>
>
Adding my 5 cents to this:

While I understand the theoretical allure of being able to use gcc on
Windows, I am very skeptical of the practical relevance.

I have worked on weird Operating Systems with odd C++ toolchains. My
experience is that you never want to base an important infrastructure
project like OpenJDK on a niche Compiler. This we call the "Little Red
Riding Hood Path" in German, where you are off the trodden paths, all alone
by yourself. You will run into many problems nobody else runs into. The
OpenJDK itself is complex enough without having to deal with an unreliable
build toolchain.

Therefore, a hypothetical "gcc on Windows" port will cause constant
maintenance efforts and be a source of fun long after you have written your
patches. Those are the immediate costs. Add to that the disturbance you
cause by invasive "shotgun spread" changes. Add to that the cost of
opportunity: your massive changes have to be reviewed by someone, and that
someone cannot review something else, possibly more important. Add to that
the effort of maintaining another build environment, which will never
replace VS, so it comes atop.

Those are the cost. If there were an entity - a corporation or foundation -
with engineers on their payroll and a good track record for keeping things
running, and that entity wanted to sponsor the "gcc on Windows" work and
shoulder at least the immediate costs, this would be worth considering. But
I don't see that.

Nor do I see a need. As long as we have a VS toolchain that works well.
Should MS ever screw that up, the arguments for a gcc port for Windows
might become more compelling. But even then, it would be worth considering
alternatives first, e.g. using Intel compilers or other products.

Lastly, I hate discouraging contributors. You are obviously very
enthusiastic and driven. If you just want to dig your teeth into OpenJDK,
there are plenty of things you could do. In general, the more focused (few
lines of code) and useful (fixing actual issues) a patch is, the better.

Cheers, Thomas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-dev/attachments/20230729/515bfc87/attachment-0001.htm>