Re: [Mingw-cross-env-list] output directory names

[Volker Grabsch]

Mark Brand <address@hidden> schrieb:
> It's not a pressing issue, but it would be a welcome simplification to  
> get rid of $(PKG)_SUBDIR.
>
> Was there originally a compelling reason for the strict treatment of the  
> subdir?

Back when we had no checksum tests, this was a good way to ensure
that the source archive contained the right package in the right
version. This, of course, doesn't apply anymore.

Another reason the use of UNPACK_PKG_ARCHIVE macro from within
the *.mk files, which is usually used for ad-hoc compiling
dependencies of native builds. This needs some more thought.

I definitely don't want to repeat the subdir-finding mechanism
with special-case handling in the *.mk files. But defining an
extra macro seems to be overkill, either.

Also, Tony introduced the repetition for the patch-applying
code to some files such as openexr.mk (when building ilmbase
inside there). This is a similar issue that we should solve.

Maybe the cleanest solution is to improve the UNPACK_PKG_ARCHIVE
macro to unpack & patch everything, and to "hard-code" the
knowledge of subdirectory names in ad-hoc builds, e.g. replacing

    cd '$(1)/$(ilmbase_SUBDIR)' && ...

with

    cd '$(1)'/ilmbase* && ...

> Regarding the references, my quick grepping of $(PKG)_SUBDIR shows that  
> it's mostly used to construct $(PKG)_FILE and more rarely $(PKG)_URL.  
> These seem easy to replace.

Those are the easy parts, indeed.

> The exotic cases are gcc, glib, openexr, postgresql and wxwidgets, some  
> of which even have references to other packages' $(PKG)_SUBDIR.  
> Alternative tricks would have to be found.

This is the "other reason" for which we sill have $(PKG)_SUBDIR,
as mentioned above. :-)


Greets,
Volker

-- 
Volker Grabsch
---<<(())>>---