Re: [RFC] [PATCH] utils.bbclass: simplify checksum check, prepare for checksums.ini removal

[Martin Jansa <martin.jansa@gmail.com>]

On Thu, Apr 08, 2010 at 09:41:37AM -0700, Chris Larson wrote:
> On Thu, Apr 8, 2010 at 6:40 AM, Martin Jansa <martin.jansa@gmail.com> wrote:
> >
> > * unify OE_STRICT_CHECKSUMS and OE_ALLOW_INSECURE_DOWNLOADS, one option
> >  for insane people should be enough, when the later is enabled, don't
> >  raise Exception even for missing md5sum/oe_sha256sum command or
> >  different checksums
> >
> 
> I find it very useful to distinguish between the missing checksum and
> invalid checksum cases.  The latter should really never be allowed, at all,
> period, imo, but the missing checksum should have an option.  If we aren't
> ready to remove the ability to allow invalid, then we need to be able to
> control the two cases independently, or via two different values in the
> variable that controls the behavior.

OK fair enough, I've expected this.. (that's also why it's RFC :)). Do
you agree that with enabled OE_ALLOW_INSECURE_DOWNLOADS it should
continue even when it's not possible to check checksum (like missing
md5sum/oe_sha256sum command)?.

> * show note, when there are checksums only in checksums.ini (prepare for
> >  script for moving all to recipes)
> >
> 
> This sounds good, though it may be something best relegated to an explicit
> sanity check, depending on how much it clutters the output.  May want to log
> it to a file like tmp/legacy-staging.log, also.
> 
> 
> > * parse checksums.ini only when there is no checksum in recipe (could be
> >  faster, but for more checked items in SRC_URI it is parsed repeatedly)
> >
> 
> "Could be" .. sounds like this isn't ready to go in yet, need to do
> profiling.  Changing something because it "could" be good is best done in
> proof of concept code, not as a part of a single patch like this one.

Both those points are assuming I'll be allowed to push "recipes: move
checksums from checksums.ini to recipes", which is being prepared ATM
with -c fetchall on all recipes. After this patch both bb.note output 
as well as parsing checksums.ini should be needed only in rare cases.

> > * if one checksum doesn't match then count and show both (md5 as well as
> >  sha256) - usefull for copy&paste checksums for new recipe.

Thanks for comments,

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa