Eric Blake wrote: ...
I propose to include the following change right after yours: The only trick is to disable the new syntax-check for hash.h, since it currently evokes a false-positive.
Is there ever a chance that #include <hash.h> in libvirt sources will pick up the wrong header? That is, should we rename one of the two files to a less-common name, then update all clients to specify which of the two versions they are grabbing? But that is independent of this patch.
This is probably a good reason to rename the gnulib one to something like gl-hash.h (probably with measures to help clients ease into the change). I'll bring it up on bug-gnulib.
Subject: [PATCH] build: update gnulib submodule to latest
* cfg.mk (local-checks-to-skip): Disable sc_prohibit_hash_without_use for now, since it fails with a false-positive match.
ACK (assuming my ACK is good for much, given my relative newcomer status).
For that matter, maybe we should push yours first; the commit message in 3/4 can then use a bit of tweaking, if we then match bootstrap to ./.gnulib/build-aux/bootstrap.
If that's all you want to change, just send me an updated patch and I'll plug it in. Before or after shouldn't matter, as long as the Makevars-removing commit and the .gnulib-udpating one are pushed at the same time.
I guess I should also do a followup patch to README-hacking to give the hint about setting GNULIB_SRCDIR to reduce download time if you already have a local copy of gnulib.git.
Good idea.