[meta-qt5] do_generate_qt_config_file not rebuilding when machine changes

[meta-qt5] do_generate_qt_config_file not rebuilding when machine changes

[Philip Craig]

Hi,

I'm using meta-qt5 with fido, building for two different machines.
Each machine can build from a clean sstate correctly, but there's a
problem with the task dependencies that means not enough is
rebuilding.

Background info: One of my systemd patches has made systemd machine
specific, so when bitbake needs to build qtbase_git, it sees a
different hash for the systemd do_populate_sysroot task and decides to
rebuild most of qtbase_git. I realise that's probably not ideal, but I
can't fix it right now.

The problem is that when bitbake starts rebuilding qtbase_git for the
second machine, it sees that the hash for do_generate_qt_config_file
is unchanged, and so it keeps using the old qt.conf, but qt.conf
contains a path to the first machine sysroot, and so it fails. More
specifically, OE_QMAKE_PATH_HOST_DATA and OE_QMAKE_PATH_HOST_LIBS both
contain STAGING_DIR_HOST.

So my question is: what is the correct way to tell bitbake that
do_generate_qt_config depends on the value of the STAGING_DIR_HOST?
Or is there some other way to fix this?

Re: [meta-qt5] do_generate_qt_config_file not rebuilding when machine changes

[Christian Ege]

Hi,


Am 8. März 2016 4:39:28 vorm. schrieb Philip Craig <philip@blackmoth.com>:

> Hi,
>
> I'm using meta-qt5 with fido, building for two different machines.
> Each machine can build from a clean sstate correctly, but there's a
> problem with the task dependencies that means not enough is
> rebuilding.
I have similar issues when building for two machines on Jethro branch.

>
> Background info: One of my systemd patches has made systemd machine
> specific, so when bitbake needs to build qtbase_git, it sees a
> different hash for the systemd do_populate_sysroot task and decides to
> rebuild most of qtbase_git. I realise that's probably not ideal, but I
> can't fix it right now.
>
> The problem is that when bitbake starts rebuilding qtbase_git for the
> second machine, it sees that the hash for do_generate_qt_config_file
> is unchanged, and so it keeps using the old qt.conf, but qt.conf
> contains a path to the first machine sysroot, and so it fails. More
> specifically, OE_QMAKE_PATH_HOST_DATA and OE_QMAKE_PATH_HOST_LIBS both
> contain STAGING_DIR_HOST.
>
> So my question is: what is the correct way to tell bitbake that
> do_generate_qt_config depends on the value of the STAGING_DIR_HOST?
> Or is there some other way to fix this?
> --
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-devel