Herbert Valerio Riedel hvriedel at
Wed Dec 3 10:18:34 UTC 2014

On 2014-12-03 at 09:48:58 +0100, Simon Peyton Jones wrote:
> I did the Ubuntu upgrade thing, and it's still happening for me too.
> I have no idea how to narrow it down some more.

I had a short conversation w/ Joachim on #ghc, and what we know so far:

For a non-failing linker_unload environment, the testprogram is linked

$ ldd tests/rts/linker_unload =>  (0x00007fff20f6c000) => /usr/lib/x86_64-linux-gnu/ (0x00007f83c5bbb000) => /lib/x86_64-linux-gnu/ (0x00007f83c58b5000) => /lib/x86_64-linux-gnu/ (0x00007f83c56ac000) => /lib/x86_64-linux-gnu/ (0x00007f83c54a8000) => /lib/x86_64-linux-gnu/ (0x00007f83c50e3000) => /lib/x86_64-linux-gnu/ (0x00007f83c4ec4000)
	/lib64/ (0x00007f83c5e76000)

Also, in the reported test failure, the failure occurs when calling
resolveObjs() on the Test.o file (after the libHS-libs were already
succesfully loadPkg()ed)

        r = loadObj(OBJPATH);
        if (!r) {
            errorBelch("loadObj(%s) failed", OBJPATH);
        r = resolveObjs();
        if (!r) {
            errorBelch("resolveObjs failed");

Alas we still don't know what property of the build-environment
determines whether this test fails or succeeds... :-/

More information about the ghc-devs mailing list