Re: gzimg hashes

On Tue, 2026-07-14 at 07:56 -0600, Brook Milligan wrote:
>=20
>=20
> > On Jul 14, 2026, at 06:30, David H. Gutteridge <[email protected]>
> > wrote:
> >=20
> > Looks like this is a parallelism issue, as I suspected. My example
> > above
> > is with "-j 8", where these checksums are spun up in parallel quite
> > quickly. So something would have to deal with this.
>=20
> Thanks for your comments.=C2=A0 I had already improved the script to use
> the build tools, but had not thought of mktemp.=C2=A0 I=E2=80=99ll fix th=
at too.
>=20
> I was initially surprised that it was not working for you, as I have
> been using this reliably for awhile, always for multiple files.=C2=A0
> Parallelism problems make sense, though.=C2=A0 I=E2=80=99m not immediatel=
y sure
> what to do, but I=E2=80=99ll work on that.
>=20

Our make(1) offers a special ".NOTPARALLEL" target that can be placed
around problem spots, so that could be one way. (I haven't tried it yet,
but I've found multiple "-j 1" builds for me always result in correct
output for the aarch64 checksum files, as you'd expect.)

Thanks,

Dave