[libvirt] [PATCH] util: Use virStorageSourceNew in virStorageFileMetadataNew

[Erik Skultety]

On Mon, Feb 18, 2019 at 08:28:54AM -0500, John Ferlan wrote:
>
>
> On 2/18/19 7:51 AM, Peter Krempa wrote:
> > On Mon, Feb 18, 2019 at 07:38:34 -0500, John Ferlan wrote:
> >>
> >>
> >> On 2/18/19 7:27 AM, Peter Krempa wrote:
> >>> Commit dcda2bf4c110 forgot to fix this one instance.
> >>>
> >>> Signed-off-by: Peter Krempa <pkrempa at redhat.com>
> >>> ---
> >>>  src/util/virstoragefile.c | 14 ++++++--------
> >>>  1 file changed, 6 insertions(+), 8 deletions(-)
> >>>
> >>
> >> Right fix as far as I'm concerned; however, I am curious to know whether
> >> this passes the MinGW build test because this is exactly where things
> >> started to break down for VIR_AUTOPTR usage.
> >
> > We'll see after I push it. I don't have mingw deployed and don't care
> > enough about that platform to do so.
> >
>
> Cannot disagree about the relevance of the importance of MinGW. Still I
> note it because it was something that caused previous changes to add
> VIR_AUTOPTR for this module to not be pushed. I was pointed in the
> direction of Andrea and the lcitool, but TBH that didn't help me much.
> Eventually I noted that Erik had run a build via some link between
> travis-ci.org and a github repo. I was able to do something similar and
> found a similar failure.

This actually passes the build on MinGW:
https://travis-ci.org/eskultety/libvirt

>
> >>
> >> VIR_AUTOUNREF could also be used more liberally in this module...
> >
> > I'll not pursue this refactor.
> >
> >>
> >>> diff --git a/src/util/virstoragefile.c b/src/util/virstoragefile.c
> >>> index 5927140a66..b2e308d81d 100644
> >>> --- a/src/util/virstoragefile.c
> >>> +++ b/src/util/virstoragefile.c
> >>> @@ -1119,22 +1119,20 @@ static virStorageSourcePtr
> >>>  virStorageFileMetadataNew(const char *path,
> >>>                            int format)
> >>>  {
> >>> -    virStorageSourcePtr def = NULL;
> >>> +    VIR_AUTOUNREF(virStorageSourcePtr) def = NULL;
> >>> +    virStorageSourcePtr ret = NULL;
> >>
> >> Erik prefers the usage of VIR_AUTO* defs last (IOW, swap these).
> >
> > Well I prefer if the returned variable is last and if the longer lines
> > are first.
> >
>
> Picking and choosing which review comments to follow is an interesting
> decision - hopefully it's not contagious.
>
> Consistency wise, VIR_AUTO* defs have been last. If it's that important
> I suppose per usual someone can come in afterwards and propose another
> patch as well as either a rule in/for make check or adjustment to the
> hacking guide.

I am obviously in favour of consistency and I'd like to us to follow it, but of'
course as a reviewer you can't really force the author to do that :/..
well, wechnically we could abuse perl once again for "the rescue" and create a
syntax-check rule, but HELL no...I agree that we might want to mention this in
the HACKING guide.

Erik