RFR: 8302460: System C++ Compiler for cross compiling gcc is dangerously wrong

Magnus Ihse Bursie ihse at openjdk.org
Wed Feb 15 14:13:08 UTC 2023


On Wed, 15 Feb 2023 13:02:18 GMT, Julian Waters <jwaters at openjdk.org> wrote:

> On my end, the uppercase `CC` ends up resolving to `gcc`, likely because `CC` doesn't exist (at least on some platforms)

I don't understand what you are saying here. Prior to you patch, we looked for exactly two executables: `CC` and `g++`, in that order. If an executable named `CC` was found in the PATH, we picked it, otherwise we looked for `g++`. If neither was found, configure would fail.

So then the question is: do you have `CC` on your path? 

You say "`CC` ends up resolving", but who is doing that resolving? Not the configure script, afaik. 

Is this, after all, some issue with case-insensitive file systems? Is this part of your attempt to get the JDK to build on Windows with a non-Microsoft toolchain? Does it mean that `cc` is found in the PATH, and that on Windows this is indistinguishable from `CC`?

Then this might still be something we want to change, but you need to rename, and rephrase, the bug report. What you want in that case is an adaptation for making a kind of cross between OS and toolchain that we do not support. I strongly suggest you keep all such changes in a separate, private branch, until you reach the point that you can actually get a JDK built, that passes a fair amount of testing. Then we can start discussing how up upstream the changes needed to do that build, if at all.

If you are experimenting like this, and find an issue that affects the existing build badly, then yes, we can fix that separately. Maybe it was your understanding that you had found such an issue, but if that is the case, I believe you are wrong.

-------------

PR: https://git.openjdk.org/jdk/pull/12558



More information about the build-dev mailing list