[Libosinfo] [PATCH libosinfo/osinfo-db 0/5] Force anchored patterns when matching regex

[Fabiano Fidêncio]

On Thu, 2018-10-04 at 11:05 +0100, Daniel P. Berrangé wrote:
> On Tue, Sep 11, 2018 at 01:15:26PM -0400, Cole Robinson wrote:
> > On 09/07/2018 03:18 PM, Fabiano Fidêncio wrote:
> > > This patch series basically consists in a fix for a possibly
> > > always
> > > present issue that I've faced Today after adding more data to
> > > osinfo-db.
> > > 
> > > Please, take a careful look at "db: Force anchored ..." patch as
> > > this is
> > > the most important patch of the series.
> > > 
> > > As this change ends up exposing a few more issues on osinfo-db,
> > > would be
> > > really nice to have the osinfo-db patches merged altogether.
> > > 
> > > It's important to note that the osinfo-db patches themselves are
> > > **not**
> > > going to break something in case they're used against a "non-
> > > patched"
> > > libosinfo.
> > > 
> > > Also, the osinfo-db patches will have their commit message edited
> > > after
> > > the libosinfo patches are pushed, so I can reference the proper
> > > commit
> > > hash in their commit messages.
> > > 
> > > libosinfo:
> > > Fabiano Fidêncio (2):
> > >    tests: Expand the arch's parser for isodetect
> > >    db: Force anchored patterns when matching regex
> > > 
> > >   osinfo/osinfo_db.c     | 2 +-
> > >   tests/test-isodetect.c | 2 ++
> > >   2 files changed, 3 insertions(+), 1 deletion(-)
> > > 
> > > osinfo-db:
> > > Fabiano Fidêncio (3):
> > >    altlinux: Fix publisher-id for 4.0 and 4.1
> > >    win10: Fix volume-id
> > >    openbsd: Fix all publisher-ids
> > > 
> > 
> > I guess technically this could constitute an API break: if a user
> > has custom
> > installed db entries and is depending on this behavior, they could
> > see
> > different results after an update. That said I doubt it's an issue
> > in
> > practice, and implying ANCHORED mode is the better long term option
> > IMO, so:
> > 
> > Reviewed-by: Cole Robinson <crobinso at redhat.com>
> 
> If we want anchored matches, then the osinfo-db regexes should
> include the "^" and "$" anchors as they see fit. We shouldn't force
> it in libosinfo API itself. Note that we explicitly allow apps to
> load & interpret the XML files directly without using libosinfo APIs,
> which further pushes to using anchors in the XML

So, your preference would be to revert the libosinfo patch and then
patch osinfo-db case-by-case?

> 
> Regards,
> Daniel