(Tinderboxed.)

(Tinderboxed.)

(Tinderboxed.)

Much weirder than that. Strace, on the AppImage I'm using for testing, shows this:

[pid  1792] mount("scribus-1.5.6.1-linux-x86_64.AppImage", "/tmp/.mount_scribuH15vRB", "fuse.scribus-1.5.6.1-linux-x86_6"..., MS_RDONLY|MS_NOSUID|MS_NODEV, "fd=5,rootmode=40000,user_id=0,gr"...) = 0

FWIW, splice: https://reviews.freebsd.org/D30597

It already suggests that's the case, though: "XXX: The following bits (from CPUID to FLAGM) were originally incorrect, but later changed to match the Linux definitions."

This looks really interesting, thanks for working on it!

(Tinderboxed.)

(Tinderboxed.)

(Tinderboxed.)

(Tinderboxed.)

Looks fine, please go ahead.

Another attempt to fix build.

Update the struct layout and defines.

(Tinderboxed.)

(Tinderboxed.)

Ah, sorry. I've removed AT_HWCAP2; the AT_PAGESZ was there before, just out of order.

Drop LINUX_AT_HWCAP2.

Hm, can you point me at the exact line?

Drop spurious diff.

(Note to self: fails tinderbox, probably depends on https://reviews.freebsd.org/D30193)

Fix tinderbox.

There's still one change to imgact_elf.c, but this one is completely Linux-specific, so I'm not sure if there's a point in splitting it off.

Rebase.

Restore flags, improve comments, style.

FWIW, splice(2) can be a wrapper around copy_file_range(2). We should also have a fallback from sendfile(2) to splice(2), for when it fails with ENOTSOCK. This is one of the things that break Mono.

Arrgh.

Sigh.

Drop unnecessary cast.

Regen and fix some issues.

(Tinderboxed.)

Sure, I'll hold off.