Re: [XEN PATCH] libs: Fix unstable libs build on FreeBSD, auto-generate version-script

[Jan Beulich <jbeulich@xxxxxxxx>]

On 15.02.2023 18:41, Anthony PERARD wrote:
> On Wed, Feb 15, 2023 at 04:30:43PM +0100, Jan Beulich wrote:
>> On 15.02.2023 16:21, Anthony PERARD wrote:
>>> @@ -13,6 +14,10 @@ MAJOR := $(shell $(XEN_ROOT)/version.sh 
>>> $(XEN_ROOT)/xen/Makefile)
>>>  endif
>>>  MINOR ?= 0
>>>  
>>> +ifeq ($(origin version-script), undefined)
>>> +version-script := libxen$(LIBNAME).map.tmp
>>> +endif
>>
>> Such a use of $(origin ...) is pretty fragile. Maybe better use ?= ?
> 
> I'm not sure why I used $(origin), I've written that more than 6 month
> ago, but this way of using it is actually described in the manual, when
> documenting ?= but with a = instead of := .
> So, maybe I wanted to have ?= but with immediate evaluation rather than
> deferred. Maybe I had issue with $(version-script) evaluating to
> different values.

What's wrong with using deferred evaluation for this macro? $(LIBNAME) (and
$(LIB_FILE_NAME)) don't vary over time, do they? Plus ...

>>>  lib$(LIB_FILE_NAME).a: $(OBJS-y)
>>
>> Seeing this right adjacent in context - any reason you use libxen$(LIBNAME)
>> and not the same lib$(LIB_FILE_NAME) for the base file name?
> 
> That was the name used before, I guess it would be fine to rename. I
> just need to set $(version-script) later as $(LIB_FILE_NAME) would not
> be defined yet.

... you'd likely deal with that issue as well, unless there's a "too early"
use of $(version-script) somewhere.

Jan