Re: [Qemu-devel] [PATCH] build: Cleanup qga make output

[Michael Roth <mdroth@linux.vnet.ibm.com>]

On 12/13/2011 08:33 AM, Adam Litke wrote:
> On Tue, Dec 13, 2011 at 10:23:18AM -0200, Luiz Capitulino wrote:
>> On Mon, 12 Dec 2011 17:38:36 -0600
>> Michael Roth<mdroth@linux.vnet.ibm.com>  wrote:
>>
>>> On 12/12/2011 05:03 PM, Anthony Liguori wrote:
>>>> On 12/07/2011 10:33 AM, Adam Litke wrote:
>>>>> Currently the make variable qapi-dir refers to the qapi-generated
>>>>> directory in
>>>>> absolute terms. This causes the harmless but ugly make output below. By
>>>>> changing this variable to the relative path the output conforms to the
>>>>> norm and
>>>>> the build works fine.
>>>>>
>>>>> Before patch:
>>>>> CC /home/aglitke/src/qemu/qapi-generated/qga-qapi-types.o
>>>>> CC /home/aglitke/src/qemu/qapi-generated/qga-qapi-visit.o
>>>>> CC /home/aglitke/src/qemu/qapi-generated/qga-qmp-marshal.o
>>>>> After patch:
>>>>> CC qapi-generated/qga-qapi-types.o
>>>>> CC qapi-generated/qga-qapi-visit.o
>>>>> CC qapi-generated/qga-qmp-marshal.o
>>>>
>>>> This was supposedly to fix a build issue that I was never able to
>>>> reproduce. I think Luiz could reproduce it though. Luiz, could you try
>>>> out Adam's patch and confirm it breaks for you?
>>
>> It doesn't :( I also tried to reproduce again with -rc4 (which was the
>> version I was getting it) but it doesn't happen anymore.
>>
>>> I think that was Stefano:
>>>
>>> http://lists.gnu.org/archive/html/qemu-devel/2011-11/msg02752.html
>>>
>>> Stefano had a patch that fixed a build breakage he saw with dirty
>>> directories by setting:
>>>
>>> qapi-dir := $(SRC_DIR)/qapi-generated
>>>
>>> That patch ended up breaking the build for others, however, and was
>>> reverted.
>>>
>>> This patch was an improvement on the original, but we'd all agreed that
>>> it wasn't necessary since we don't support working around issues related
>>> to dirty directories.
>
> I am lacking some context here.  I am confused how a dirty directory could
> change the output in this way.

I guess I'm lacking context as well, I wasn't aware that Luiz had also 
hit an issue related to this patch, was just going off what I'd read in 
the thread where I posted this patch...

In Stefano's case, he'd noticed his build issue went away if

a) this (rather, Stefano's original) patch was applied, or
b) if he did a `make distclean` in his source directory beforehand

So that's why we assumed dirty directories were the problem.

I'm not sure of the circumstances behind the breakage Luiz saw, so I'd 
err on the side of caution and keep this patch until that's clarified.

>
>> What kind of issue? I'm 100% certain that my tree was clean, the only
>> thing I did not do was to clone it again.
>>
>> We have had races when building the qapi in the past, maybe that's the
>> problem and commit 9b12940 makes it less likely.
>>
>> We could revert it and wait for the problem to happen again...
>
> I would be fine with doing whatever is least likely to cause anyone else
> problems.  I wouldn't want to break the actual build just to make the output
> look nice.
>
>>
>>>
>>> So if it's ugly, we can safely drop it.
>>>
>>>>
>>>> Regards,
>>>>
>>>> Anthony Liguori
>>>>
>>>>>
>>>>> Signed-off-by: Adam Litke<agl@us.ibm.com>
>>>>> ---
>>>>> Makefile | 2 +-
>>>>> 1 files changed, 1 insertions(+), 1 deletions(-)
>>>>>
>>>>> diff --git a/Makefile b/Makefile
>>>>> index 301c75e..7c93739 100644
>>>>> --- a/Makefile
>>>>> +++ b/Makefile
>>>>> @@ -168,7 +168,7 @@ check-qjson: check-qjson.o $(qobject-obj-y)
>>>>> $(tools-obj-y)
>>>>> test-coroutine: test-coroutine.o qemu-timer-common.o async.o
>>>>> $(coroutine-obj-y) $(tools-obj-y)
>>>>>
>>>>> $(qapi-obj-y): $(GENERATED_HEADERS)
>>>>> -qapi-dir := $(BUILD_DIR)/qapi-generated
>>>>> +qapi-dir := qapi-generated
>>>>> test-visitor.o test-qmp-commands.o qemu-ga$(EXESUF): QEMU_CFLAGS += -I
>>>>> $(qapi-dir)
>>>>> qemu-ga$(EXESUF): LIBS = $(LIBS_QGA)
>>>>>
>>>>
>>>>
>>>
>>
>