On 03/07/2018 09:39 AM, Richard W.M. Jones wrote:

Needs rebasing now that the test no longer relies on file-data.
Also, I still appreciate being able to override $QEMU_IO from the
command line (to point to an alternative version on the fly), so
even though I agree with your Makefile changes, I disagree with
dropping the defaulting of QEMU_IO to qemu-io,

Does setting $PATH not work for this purpose?


  # Populate file, and sanity check that qemu-io can issue parallel requests
  printf '%1024s' . > test-parallel-file.data
-$QEMU_IO -f raw -c "aio_write -P 1 0 512" -c "aio_write -P 2 512 512" \
+qemu-io -f raw -c "aio_write -P 1 0 512" -c "aio_write -P 2 512 512" \

as well as disagree with hard-coding only the first qemu-io in $PATH
here, instead of allowing a command-line override.

Well, using PATH overrides might work, but it can also be more verbose. It's also trickier: if you have two binaries residing in a single directory - say qemu-io and qemu-img - but only want the override to apply for just one of those binaries, then a direct QEMU_IO override would do it at once, but a PATH override would require creating a temporary directory, adding a symlink in the temp directory to pick up the one binary that will be used as an override, then pointing PATH to pick up the temp directory first.

But since you are right that a PATH override is always possible, even if more verbose, I won't insist on a QEMU_IO override as it is merely syntactic sugar.

Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org

Libguestfs mailing list

Reply via email to