On Mon, Mar 22, 2010 at 10:54:09AM -0600, Eric Blake wrote:
On 03/22/2010 09:03 AM, Daniel P. Berrange wrote:
On Mon, Mar 22, 2010 at 02:25:43AM +0100, Matthias Bolte wrote:
sscanf doesn't support the L modifier on Windows and gnulib has no replacement for the scanf functions. Just replace the function with a stub on Windows, because it's not used on the libvirt client side. --- src/util/pci.c | 14 ++++++++++++++ 1 files changed, 14 insertions(+), 0 deletions(-)
We already had this problem with printf(). For that gnulib provided us with a replacement that worked. We should probably pull in the scanf module from gnulib for equivalent compatability. Even though this code isn't technically required, other places may start using scanf & trip up on this problem
But that's the point that Matthias made - currently, gnulib does NOT provide a scanf module. Why? Because scanf comes with its own set of usability pitfalls (scanf("%d",&int) cannot report whether integer overflow occurred), so no one has made it a high enough priority to start replacing the portability pitfalls.
Oh, I mis-read the original description!
I've already mentioned that it would be a better cleanup to stop using *scanf altogether; but that would be an independent cleanup, unrelated to this particular patch.
For this particular patch, mingw also lacks /proc/iomem, so the fopen earlier in pciWaitForDeviceCleanup should have already failed before we ever get to the problematic sscanf("%Lx"). Therefore, do we even need this patch?
I don't think its high priority to merge this, since its not an actual compile failure - I agree we'd be better to just kill off the use of scanf() altogether Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://deltacloud.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|