Re: [PATCH] add top-level .gitattributes file, dealing with .depex

[Leif Lindholm <leif.lindholm@linaro.org>]

On Mon, Aug 01, 2016 at 12:03:06AM -0700, Jordan Justen wrote:
> On 2016-07-31 16:52:23, Kinney, Michael D wrote:
> > Jordan,
> > 
> > UEFI Drivers distributed as binaries do not need depex sections.
> > 
> > PI modules distributed as binaries do require a .depex binary.
> > 
> 
> They may require a depex, but, as mentioned below, they can also add
> it directly in the .fdf. As it stands, apparently we have 1 .depex
> file in the tree, and it is unused.
> 
> Aside from this, under what conditions would we take such binaries
> into the EDK II tree? Today we have the ShellPkg and FatPkg binaries
> in the EDK II tree, but we recently discussed removing even those.

While I don't disagree, the PI dependency expression instruction set
(section 10.7, PI spec 1.4 vol2) does not look Turing complete to me.
Meaning it's "binary" in much the same way a .uni file is.

(This is historically where someone pulls out an operating system
kernel written entirely in PI depex binary.)

> For an open source project, I think it is best to not have pre-built
> binaries, unless there is some very compelling reason. Previously
> there was some license funniness on FatPkg, but now that is gone. If
> it took an hour to build FatPkg, then that might also be something to
> discuss. :)
>
> I don't think adding the .gitattributes is really a problem, aside
> from the fact that it implies that we might actually have a reason to
> add a .depex file to the source tree.

And I agree it would send that signal.

Regards,

Leif

> -Jordan
> 
> > So I would recommend .depex binary files be treated the same as 
> > binary .efi files by GIT.  So it does sound like we need some
> > minor updates to GIT attributes.
> > 
> > If we have an example of a binary module that is providing more
> > binary leaf sections than are actually required and/or used, then
> > yes, the binary module should be cleaned up to remove the unused
> > content.
> > 
> > Thanks,
> > 
> > Mike
> > 
> > > -----Original Message-----
> > > From: Justen, Jordan L
> > > Sent: Sunday, July 31, 2016 3:58 PM
> > > To: Leif Lindholm <leif.lindholm@linaro.org>; Tim Lewis <tim.lewis@insyde.com>
> > > Cc: Laszlo Ersek <lersek@redhat.com>; Kinney, Michael D <michael.d.kinney@intel.com>;
> > > edk2-devel@ml01.01.org <edk2-devel@lists.01.org>; Andrew Fish <afish@apple.com>
> > > Subject: Re: [edk2] [PATCH] add top-level .gitattributes file, dealing with .depex
> > > 
> > > On 2016-07-30 11:33:43, Leif Lindholm wrote:
> > > > Hi Tim,
> > > >
> > > > Thanks for the warning, and investigation.
> > > >
> > > > Does this mean that you think we should ban the inclusion of .depex
> > > > files in EDK2, including future platform trees?
> > > 
> > > I don't know about banning it, but at least we could wait for someone
> > > to make a reasonable argument why they are needed.
> > > 
> > > Even for binary only modules, it looks like the fdf method outlined
> > > below is preferable to a pre-built .depex.
> > > 
> > > If (at a future point) the reason for using a .depex is to support a
> > > binary only module in a supposedly open platform under EDK II, then I
> > > guess we can decide if that is a good idea at that point.
> > > 
> > > Should we delete this one unused .depex from the tree?
> > > 
> > > -Jordan
> > > 
> > > > (If not, this patch is
> > > > still needed for git to work predictably with these files.)
> > > >
> > > > Regards,
> > > >
> > > > Leif
> > > >
> > > > On Fri, Jul 29, 2016 at 05:12:49PM +0000, Tim Lewis wrote:
> > > > > It appears that this file is not actually used. It is only
> > > > > referenced in the [Rule.Common.UEFI_DRIVER.NATIVE_BINARY] rule in
> > > > > PlatformPkg.fdf. A little further research shows that an alternate
> > > > > method was used for the actual GOP binary (see below). A grep of the
> > > > > entire tree shows that no one uses this rule NATIVE_BINARY. So it
> > > > > looks like it can just be cut out.
> > > > >
> > > > > BTW, the downside of the method used for the binary version of the
> > > > > GOP driver, is that those drivers cannot use PCDs, since the PCD
> > > > > database is created based on references in the .inf. GOP works
> > > > > because it is pure UEFI and (therefore) doesn't use PCDs.
> > > > >
> > > > > Tim
> > > > >
> > > > > FILE DRIVER = FF0C8745-3270-4439-B74F-3E45F8C77064 {
> > > > >   SECTION DXE_DEPEX_EXP = {gPlatformGOPPolicyGuid}
> > > > >   SECTION PE32 =
> > > Vlv2MiscBinariesPkg/GOP/7.2.1011/RELEASE_VS2008x86/$(DXE_ARCHITECTURE)/IntelGopDriver.e
> > > fi
> > > > >   SECTION UI = "IntelGopDriver"
> > > > > }
> > > > >
> > > > > -----Original Message-----
> > > > > From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On Behalf Of Leif
> > > Lindholm
> > > > > Sent: Friday, July 29, 2016 9:45 AM
> > > > > To: Laszlo Ersek <lersek@redhat.com>
> > > > > Cc: michael.d.kinney@intel.com; Jordan Justen <jordan.l.justen@intel.com>; edk2-
> > > devel@ml01.01.org; Andrew Fish <afish@apple.com>
> > > > > Subject: Re: [edk2] [PATCH] add top-level .gitattributes file, dealing with .depex
> > > > >
> > > > > On Thu, Jul 07, 2016 at 05:03:13PM +0200, Laszlo Ersek wrote:
> > > > > > On 07/07/16 16:24, Leif Lindholm wrote:
> > > > > > > Git tends to see .depex files as text, causing hideous patches being
> > > > > > > generated (and breaking PatchCheck.py).
> > > > > > >
> > > > > > > Add a .gitattributes file instructing git to treat them as binary.
> > > > > > >
> > > > > > > Contributed-under: TianoCore Contribution Agreement 1.0
> > > > > > > Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
> > > > > > > ---
> > > > > > >  .gitattributes | 1 +
> > > > > > >  1 file changed, 1 insertion(+)
> > > > > > >  create mode 100644 .gitattributes
> > > > > > >
> > > > > > > diff --git a/.gitattributes b/.gitattributes new file mode 100644
> > > > > > > index 0000000..2d8a45b
> > > > > > > --- /dev/null
> > > > > > > +++ b/.gitattributes
> > > > > > > @@ -0,0 +1 @@
> > > > > > > +*.depex binary
> > > > > > >
> > > > > >
> > > > > > What generates .depex files? I've never seen any.
> > > > > >
> > > > > > Also, unless you add .depex files with "git add" to the set of tracked
> > > > > > files, no patches / diffs should cover them. What am I missing? :)
> > > > > >
> > > > > > ... Hm, after
> > > > > >
> > > > > > $ find . -iname "*.depex"
> > > > > >
> > > > > > I see .depex files in Build/ (which should be ignored altogether), and
> > > > > >
> > > > > > ./Vlv2TbltDevicePkg/IntelGopDepex/IntelGopDriver.depex
> > > > > >
> > > > > > Why does that file exist in the tree? Let me see... git log says nothing relevant
> > > (the file dates back to commit 3cbfba02fef9, "Upload BSD-licensed Vlv2TbltDevicePkg and
> > > Vlv2DeviceRefCodePkg to").
> > > > > >
> > > > > > Grepping the tree for the filename itself leads to:
> > > > > >
> > > > > > Vlv2TbltDevicePkg/PlatformPkg.fdf:    DXE_DEPEX DXE_DEPEX Optional
> > > $(WORKSPACE)/$(PLATFORM_PACKAGE)/IntelGopDepex/IntelGopDriver.depex
> > > > > > Vlv2TbltDevicePkg/PlatformPkgGcc.fdf:    DXE_DEPEX DXE_DEPEX Optional
> > > $(WORKSPACE)/$(PLATFORM_PACKAGE)/IntelGopDepex/IntelGopDriver.depex
> > > > > >
> > > > > > Do these rules exist to override the DEPEX sections of binary-only modules? If
> > > so: that's horrible.
> > > > > >
> > > > > > Anyway, given that edk2 contains at least one .depex file, and your patch is
> > > correct according to <https://git-scm.com/book/en/v2/Customizing-Git-Git-Attributes>:
> > > > > >
> > > > > > Reviewed-by: Laszlo Ersek <lersek@redhat.com>
> > > > >
> > > > > Thanks!
> > > > >
> > > > > I had hoped for comments from someone else on cc, since we don't have any
> > > Maintainers.txt entry for the top level directory :)
> > > > >
> > > > > But if I don't hear anything before Monday, I'll push it then.
> > > > >
> > > > > Regards,
> > > > >
> > > > > Leif
> > > > >
> > > > > _______________________________________________
> > > > > edk2-devel mailing list
> > > > > edk2-devel@lists.01.org
> > > > > https://lists.01.org/mailman/listinfo/edk2-devel