[PATCH 11/11] Remove including timestamp.h in version.h

Pali Rohár pali at kernel.org
Wed Aug 4 23:46:53 CEST 2021


On Wednesday 04 August 2021 10:40:42 Tom Rini wrote:
> On Wed, Aug 04, 2021 at 08:36:27AM -0600, Simon Glass wrote:
> > Hi Tom,
> > 
> > On Wed, 4 Aug 2021 at 06:44, Tom Rini <trini at konsulko.com> wrote:
> > >
> > > On Wed, Aug 04, 2021 at 07:59:21AM +0200, Wolfgang Denk wrote:
> > >
> > > > Dear Tom,
> > > >
> > > > In message <20210802213100.GG9379 at bill-the-cat> you wrote:
> > > > >
> > > > > I think that if nothing changes the banner not changing is the right
> > > > > behavior.
> > > >
> > > > Hm... is it?  How about "external" changes like building with a
> > > > different tool chain? Then at least the "version" command should
> > > > produce a corresponding output, which means some parts _have_ to be
> > > > recompiled, resulting in a new banner, too ?
> > >
> > > There's some misunderstanding about what this series does, and does not
> > > do.  First, with the series applied and you don't force
> > > SOURCE_DATE_EPOCH, and you run make, and you re-run make, cmd/version.c
> > > is rebuilt (but only that and not a bunch of other things anymore) and
> > > the banner timestamp changes.  Second, if you do set SOURCE_DATE_EPOCH
> > > to a fixed value (so you're trying to reproduce a build with a specific
> > > timestamp) AND re-run make, nothing changes, as it should.  Third, if
> > > you change gcc, everything gets rebuilt including cmd/version.c and a
> > > new banner.
> > 
> > Yes that's my understanding and it seems fine.
> 
> To be clear, I did a quick test with the series (git pw is very handy)
> for the above.

Thanks for clarification! You are of course right.


More information about the U-Boot mailing list