[PATCH RESEND] bitbake.conf: do not let DISTROOVERRIDES fall back to empty string

[PATCH RESEND] bitbake.conf: do not let DISTROOVERRIDES fall back to empty string

[Masahiro Yamada]

Commit abacd7243a2e ("bitbake.conf: fix distro-less failure expanding
FILESPATH") mended the littered DISROOVERRIDES when DISTRO is unset.

Some time later, commit b7279f996397 ("bitbake.conf: Default DISTRO to
nodistro") added a weak default to 'DISTRO'.

So, this line can go back to the original simple definition.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
---

 meta/conf/bitbake.conf | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
index 4b544a22cd..eb3dd0007e 100644
--- a/meta/conf/bitbake.conf
+++ b/meta/conf/bitbake.conf
@@ -739,7 +739,7 @@ DISTRO_NAME ??= "OpenEmbedded"
 OVERRIDES = "${TARGET_OS}:${TRANSLATED_TARGET_ARCH}:pn-${PN}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}:${CLASSOVERRIDE}${LIBCOVERRIDE}:forcevariable"
 LIBCOVERRIDE ?= ""
 CLASSOVERRIDE ?= "class-target"
-DISTROOVERRIDES ?= "${@d.getVar('DISTRO') or ''}"
+DISTROOVERRIDES ?= "${DISTRO}"
 MACHINEOVERRIDES ?= "${MACHINE}"
 
 FILESOVERRIDES = "${TRANSLATED_TARGET_ARCH}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}"
-- 
2.25.1

Re: [PATCH RESEND] bitbake.conf: do not let DISTROOVERRIDES fall back to empty string

[Richard Purdie]

On Sun, 2020-04-26 at 00:19 +0900, Masahiro Yamada wrote:
> Commit abacd7243a2e ("bitbake.conf: fix distro-less failure expanding
> FILESPATH") mended the littered DISROOVERRIDES when DISTRO is unset.
> 
> Some time later, commit b7279f996397 ("bitbake.conf: Default DISTRO
> to
> nodistro") added a weak default to 'DISTRO'.
> 
> So, this line can go back to the original simple definition.
> 
> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
> ---
> 
>  meta/conf/bitbake.conf | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
> index 4b544a22cd..eb3dd0007e 100644
> --- a/meta/conf/bitbake.conf
> +++ b/meta/conf/bitbake.conf
> @@ -739,7 +739,7 @@ DISTRO_NAME ??= "OpenEmbedded"
>  OVERRIDES = "${TARGET_OS}:${TRANSLATED_TARGET_ARCH}:pn-
> ${PN}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}:${CLASSOVERRIDE}${LIBCOV
> ERRIDE}:forcevariable"
>  LIBCOVERRIDE ?= ""
>  CLASSOVERRIDE ?= "class-target"
> -DISTROOVERRIDES ?= "${@d.getVar('DISTRO') or ''}"
> +DISTROOVERRIDES ?= "${DISTRO}"
>  MACHINEOVERRIDES ?= "${MACHINE}"

Whilst this can't happen with our standard definitions now, the side
effects of it not being set are nasty so I'd prefer to retain this.

Cheers,

Richard

Re: [OE-core] [PATCH RESEND] bitbake.conf: do not let DISTROOVERRIDES fall back to empty string

[Andre McCurdy]

On Sat, Apr 25, 2020 at 8:38 AM Richard Purdie
<richard.purdie@linuxfoundation.org> wrote:
> On Sun, 2020-04-26 at 00:19 +0900, Masahiro Yamada wrote:
> > Commit abacd7243a2e ("bitbake.conf: fix distro-less failure expanding
> > FILESPATH") mended the littered DISROOVERRIDES when DISTRO is unset.
> >
> > Some time later, commit b7279f996397 ("bitbake.conf: Default DISTRO
> > to
> > nodistro") added a weak default to 'DISTRO'.
> >
> > So, this line can go back to the original simple definition.
> >
> > Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
> > ---
> >
> >  meta/conf/bitbake.conf | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
> > index 4b544a22cd..eb3dd0007e 100644
> > --- a/meta/conf/bitbake.conf
> > +++ b/meta/conf/bitbake.conf
> > @@ -739,7 +739,7 @@ DISTRO_NAME ??= "OpenEmbedded"
> >  OVERRIDES = "${TARGET_OS}:${TRANSLATED_TARGET_ARCH}:pn-
> > ${PN}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}:${CLASSOVERRIDE}${LIBCOV
> > ERRIDE}:forcevariable"
> >  LIBCOVERRIDE ?= ""
> >  CLASSOVERRIDE ?= "class-target"
> > -DISTROOVERRIDES ?= "${@d.getVar('DISTRO') or ''}"
> > +DISTROOVERRIDES ?= "${DISTRO}"
> >  MACHINEOVERRIDES ?= "${MACHINE}"
>
> Whilst this can't happen with our standard definitions now, the side
> effects of it not being set are nasty so I'd prefer to retain this.

What exactly can go wrong? MACHINEOVERRIDES is empty every time we
build a -native recipe, so just having :: in OVERRIDES or
FILESOVERRIDES doesn't appear to be an issue. Is there something
special about DISTROOVERRIDES ?

Re: [OE-core] [PATCH RESEND] bitbake.conf: do not let DISTROOVERRIDES fall back to empty string

[Richard Purdie]

On Sat, 2020-04-25 at 14:37 -0700, Andre McCurdy wrote:
> On Sat, Apr 25, 2020 at 8:38 AM Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> > On Sun, 2020-04-26 at 00:19 +0900, Masahiro Yamada wrote:
> > > diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
> > > index 4b544a22cd..eb3dd0007e 100644
> > > --- a/meta/conf/bitbake.conf
> > > +++ b/meta/conf/bitbake.conf
> > > @@ -739,7 +739,7 @@ DISTRO_NAME ??= "OpenEmbedded"
> > >  OVERRIDES = "${TARGET_OS}:${TRANSLATED_TARGET_ARCH}:pn-
> > > ${PN}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}:${CLASSOVERRIDE}${LI
> > > BCOV
> > > ERRIDE}:forcevariable"
> > >  LIBCOVERRIDE ?= ""
> > >  CLASSOVERRIDE ?= "class-target"
> > > -DISTROOVERRIDES ?= "${@d.getVar('DISTRO') or ''}"
> > > +DISTROOVERRIDES ?= "${DISTRO}"
> > >  MACHINEOVERRIDES ?= "${MACHINE}"
> > 
> > Whilst this can't happen with our standard definitions now, the
> > side
> > effects of it not being set are nasty so I'd prefer to retain this.
> 
> What exactly can go wrong? MACHINEOVERRIDES is empty every time we
> build a -native recipe, so just having :: in OVERRIDES or
> FILESOVERRIDES doesn't appear to be an issue. Is there something
> special about DISTROOVERRIDES ?

native.bbclass sets MACHINEOVERRIDES = "" whereas with this change
DISTROOVERRIDES would be set to "${DISTRO}" instead of "" as it
currently is today. I think unexpanded variables were sub-optimal in
FILESOVERRIDES but I don't remember the details of why. 

Cheers,

Richard

Re: [OE-core] [PATCH RESEND] bitbake.conf: do not let DISTROOVERRIDES fall back to empty string

[Masahiro Yamada]

On Sun, Apr 26, 2020 at 5:13 PM Richard Purdie
<richard.purdie@linuxfoundation.org> wrote:
>
> On Sat, 2020-04-25 at 14:37 -0700, Andre McCurdy wrote:
> > On Sat, Apr 25, 2020 at 8:38 AM Richard Purdie
> > <richard.purdie@linuxfoundation.org> wrote:
> > > On Sun, 2020-04-26 at 00:19 +0900, Masahiro Yamada wrote:
> > > > diff --git a/meta/conf/bitbake.conf b/meta/conf/bitbake.conf
> > > > index 4b544a22cd..eb3dd0007e 100644
> > > > --- a/meta/conf/bitbake.conf
> > > > +++ b/meta/conf/bitbake.conf
> > > > @@ -739,7 +739,7 @@ DISTRO_NAME ??= "OpenEmbedded"
> > > >  OVERRIDES = "${TARGET_OS}:${TRANSLATED_TARGET_ARCH}:pn-
> > > > ${PN}:${MACHINEOVERRIDES}:${DISTROOVERRIDES}:${CLASSOVERRIDE}${LI
> > > > BCOV
> > > > ERRIDE}:forcevariable"
> > > >  LIBCOVERRIDE ?= ""
> > > >  CLASSOVERRIDE ?= "class-target"
> > > > -DISTROOVERRIDES ?= "${@d.getVar('DISTRO') or ''}"
> > > > +DISTROOVERRIDES ?= "${DISTRO}"
> > > >  MACHINEOVERRIDES ?= "${MACHINE}"
> > >
> > > Whilst this can't happen with our standard definitions now, the
> > > side
> > > effects of it not being set are nasty so I'd prefer to retain this.
> >
> > What exactly can go wrong? MACHINEOVERRIDES is empty every time we
> > build a -native recipe, so just having :: in OVERRIDES or
> > FILESOVERRIDES doesn't appear to be an issue. Is there something
> > special about DISTROOVERRIDES ?
>
> native.bbclass sets MACHINEOVERRIDES = "" whereas with this change
> DISTROOVERRIDES would be set to "${DISTRO}" instead of "" as it
> currently is today. I think unexpanded variables were sub-optimal in
> FILESOVERRIDES but I don't remember the details of why.


Now, the same file defines this:

DISTRO ??= "nodistro"


In which case is "${DISTRO}" left unexpanded?



-- 
Best Regards
Masahiro Yamada