Re: [Qemu-devel] [PATCH v2 1/4] build: allow setting a custom GIT binary for transparent proxying

[Alexey Kardashevskiy]

On 30/10/17 18:52, Daniel P. Berrange wrote:
> On Mon, Oct 30, 2017 at 10:49:01AM +1100, Alexey Kardashevskiy wrote:
>> On 30/10/17 03:29, Daniel P. Berrange wrote:
>>> On Mon, Oct 30, 2017 at 01:08:56AM +1100, Alexey Kardashevskiy wrote:
>>>>
>>>> I run ./scripts/git-submodule.sh on a server (where source directory is
>>>> writeable), them "rm .git-submodule-status", then run "Makefile" on a build
>>>> machine (which has the source directory mounted as read-only). I basically
>>>> recreate the situation where I was when I just discovered this brand new
>>>> ./scripts/git-submodule.sh.
>>>
>>> Don't rm the .git-submodule-status. That means 'make' thinks the modules
>>> are out of date.
>>>
>>> Just run 'scripts/git-submodules.sh ...mods..' on the writable source
>>> dir, and then run 'make' on the build machine.
>>>
>>>> I know that now, all I am asking is an error message to print exact command
>>>> to run...
>>>
>>> If you hadn't deleted the .git-submodule-status, it would have worked fine.
>>
>> No.
>>
>> I do this on a server:
>>
>> [vpl1 qemu]$ git co v2.10.0
>> At this point no .git-submodule-status is expected/required/exist.
>>
>> [vpl1 qemu]$ ssh aikhostos2 make -C /home/aik/pbuild/qemu-aikhostos2-ppc64/
>> -j24
>>
>> Configure succeeds, compiles just fine, it has been working like this for
>> years.
>>
>>
>> Now:
>>
>> [vpl1 qemu]$ git co git-submodule  (this is your stuff)
>> [vpl1 qemu]$ ssh aikhostos2 make -C /home/aik/pbuild/qemu-aikhostos2-ppc64/
>> -j24
>> touch: cannot touch ‘.git-submodule-status’: Read-only file system
>> make: Entering directory `/home/aik/pbuild/qemu-aikhostos2-ppc64'
>>   GEN     ppc64-softmmu/config-devices.mak.tmp
>> [...]
>>
>> ./scripts/git-submodule.sh: line 74: .git-submodule-status: Read-only file
>> system
>> ./scripts/git-submodule.sh: failed to save git submodule status
>> make: *** [git-submodule-update] Error 1
> 
> Oh I see, the git submodules list is empty because you have a previous
> built source tree, so its shortcircuiting the extra check I added. Tihs
> is easy enough to address


btw why is the name "git-submodule.sh", not update-submodule.sh or
update-git-submodule.sh on a par with update-linux-headers.sh?




-- 
Alexey