[Buildroot] <package>_VER vs <package>_VERSION

[Ulf Samuelsson <ulf@atmel.com>]

>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>> 
>>> Ulf Samuelsson wrote:
>>> > Some packages use <package>_VER  and some use  <package>_VERSION
>>> > and some does not use this at all and hardcodes the version.
>>> > 
>>> > Suggest we clean this up so we use a single template.
>>> > 
>>> > Suggest we do it using the following template, allowing user to override.
>>> > 
>>> > ifeq    ($(<package>_VERSION),)
>>> > <package>_VERSION:=X.Y.Z
>>> > endif
>>> > 
>>> That's an interesting idea. I agree on the _VER and _VERSION clean-up. How
>>> about you submit a patch for that first? Thanks.
>>> 
>>> - -Steve
>>
>>Some packages do not have a version:
>>customize
> 
> I'd remove that one, but i guess you're opposed to that ;)

It will be OK to remove if I create <buildroot>/local !

>>gnuconfig
> 
> That's just CVS (nothing interresting was changed there, i looked about
> three weeks ago).
> 
>>hostap Always use a snapshot
>>microcom This is not clean, version = 1.02, 
>> package is m102.tar.gz
> 
> "clean", well, live with it until an upstream maintainer fixes this
> cosmetic issue.

Don't think it is maintained any longer...
We might as well "maintain" it ourself by applying the 
existing patches and put on a server somewhere
and call it 1.03 or whatever.

>>jpeg    6b vs v6b
> 
> about the same as microcom
> 

There are a number of patches for AVR32, so again
we might want to create our "own" version at some point
in the future.

>>mtd Configurable from menuconfig or snapshot

>>sfdisk No version, this is located on www.uclibc.org so we 
>> should change to sfdisk-1.00.tar.bz2
> 
> 1.00 is inappropriate. Either use the official version from util-linux
> or implement a _small_ sfdisk in busybox (but we already have fdisk,
> IIRC) or leave it alone, IMHO.
> 

We could call it 0.01 for all I care.
Have no clue what it does, but I think we should strive to
have version numbers for all supported packages.
Maybe this should just be removed from buildroot?

>>tinylogin Makefile force use of snapshot,
>> Configure using menuconfig?
> 
> standalone tinylogin is deprecated and unmaintained from my POV (use
> busybox which is actively maintained).
> 

Remove?

>>ttcp No version
> 
> "No version" depends on you view. It clearly reads
> "ttcp.c $Revision: 1.12 $";
> 

Just talking about names of files and tarballs for the buildroot Makefile.
There is no tarball for TTCP, you just download a file ttcp.c.
which does not have a version number.
I.E: it it not called ttcp-1.12.c

>>Index: package/qte/qte.mk
>>===================================================================
>>--- package/qte/qte.mk (revision 19021)
>>+++ package/qte/qte.mk (arbetskopia)
>>@@ -222,18 +222,18 @@
>> 
>> $(QTE_QTE_LIB): $(QTE_QTE_DIR)/src-mt.mk
>> export QTDIR=$(QTE_QTE_DIR); export QPEDIR=$(QTE_QTOPIA_DIR); export
>>PATH=$(STAGING_DIR)/bin:$$QTDIR/bin:$$PATH; \
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) $(TARGET_CC) -C $(QTE_QTE_DIR) src-mt
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) $(TARGET_CC) DESTDIR=
>>$(TARGET_DIR)/lib -C $(QTE_QTE_DIR) src-mt
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) $(TARGET_CC) -C $(QTE_QTE_DIR) src-mt
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) $(TARGET_CC) DESTDIR=
> 
> superfluous TARGET_CC
> 
>>$(TARGET_DIR)/lib -C $(QTE_QTE_DIR) src-mt
>> # ... and make sure it actually built... grrr... make deep-deep-deep
>>makefile recursion for this habit
>> test -f $@
>> 
>> $(QTE_QTOPIA_FILE): $(QTE_QTOPIA_DIR)/.configured
>> export QTDIR=$(QTE_QT3_DIR); export QPEDIR=$(QTE_QTOPIA_DIR); export
>>PATH=$(STAGING_DIR)/bin:$$QTDIR/bin:$$PATH; \
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) CC=$(TARGET_CC) -C $(QTE_QTOPIA_DIR)
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(QTE_QTOPIA_DIR)
>> 
>> $(QTE_QTOPIA_IFILE): $(QTE_QTOPIA_FILE)
>> export QTDIR=$(QTE_QT3_DIR); export QPEDIR=$(QTE_QTOPIA_DIR); export
>>PATH=$(STAGING_DIR)/bin:$$QTDIR/bin:$$PATH; \
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) CC=$(TARGET_CC) -C $(QTE_QTOPIA_DIR)
>>install PREFIX=$(TARGET_DIR)
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(QTE_QTOPIA_DIR) install PREFIX=
>>$(TARGET_DIR)
>> 
>> 
>> qte:: $(QTE_QTE_LIB)
>>Index: package/tinyx/tinyx.mk
>>===================================================================
>>--- package/tinyx/tinyx.mk (revision 19021)
>>+++ package/tinyx/tinyx.mk (arbetskopia)
>>@@ -85,7 +85,7 @@
>> rm -f $(TINYX_BINX)/Xfbdev
>> #make World CROSSCOMPILEFLAGS="CROSSCOMPILEDIR=<cross compiler dir>";
>> #( cd $(TINYX_DIR) ; $(MAKE) World CROSSCOMPILEFLAGS="CROSSCOMPILEDIR=
>>$(STAGING_DIR)/bin" )
>>- #( cd $(TINYX_DIR) ; $(TARGET_CONFIGURE_OPTS) $(MAKE) World )
>>+ #( cd $(TINYX_DIR) ; $(MAKE) $(TARGET_CONFIGURE_OPTS) World )
> 
> well, yea, but i don't expect that you got a real error from this hunk
> ;)

No, I am not changing this due to error I detected.
If it is wrong to use $(TARGET_CONFIGURE_OPTS) $(MAKE) for one package
then it should be wrong to use that for every other package.

Even if it does not hurt this package, someone might use a *.mk as an example
when generating support for a new package, and it would be good to remove
all bad examples, even if they are commented away like above.

> 
>> #
>> #mv $(TINYX_DIR)/Makefile $(TINYX_DIR)/Makefile.xxxx
>> #echo "AS=$(TARGET_CROSS)as" > $(TINYX_DIR)/Makefile
>>Index: package/samba/samba.mk
>>===================================================================
>>--- package/samba/samba.mk (revision 19021)
>>+++ package/samba/samba.mk (arbetskopia)
>>@@ -46,7 +46,7 @@
>> touch $(SAMBA_DIR)/.configured
>> 
>> $(SAMBA_DIR)/$(SAMBA_BINARY): $(SAMBA_DIR)/.configured
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) CC=$(TARGET_CC) -C $(SAMBA_DIR)
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(SAMBA_DIR)
>> 
>> SAMBA_TARGETS_ :=
>> SAMBA_TARGETS_y :=
>>@@ -79,7 +79,7 @@
>> SAMBA_TARGETS_$(BR2_PACKAGE_SAMBA_WBINFO) += usr/bin/wbinfo
>> 
>> $(TARGET_DIR)/$(SAMBA_TARGET_BINARY): $(SAMBA_DIR)/$(SAMBA_BINARY)
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) CC=$(TARGET_CC) \
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) \
>> prefix="${TARGET_DIR}/usr" \
>> BASEDIR="${TARGET_DIR}/usr" \
>> SBINDIR="${TARGET_DIR}/usr/sbin" \
> 
> samba is an excellent example where (it uses autoconf) i expect all
> these duplication to be superfluous; configure should have set up this
> stuff properly, didn't it? (didn't look)
> 



>>--- package/libevent/libevent.mk (revision 19021)
>>+++ package/libevent/libevent.mk (arbetskopia)
>>@@ -37,7 +37,7 @@
>> touch $@
>> 
>> $(LIBEVENT_DIR)/$(LIBEVENT_BINARY): $(LIBEVENT_DIR)/.configured
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) CC=$(TARGET_CC) -C $(LIBEVENT_DIR)
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(LIBEVENT_DIR)
>> 
>> $(STAGING_DIR)/$(LIBEVENT_TARGET_BINARY):
>>$(LIBEVENT_DIR)/$(LIBEVENT_BINARY)
>> $(MAKE) -C $(LIBEVENT_DIR) DESTDIR=$(STAGING_DIR) install
> 
> I don't use this package but installing to STAGING_DIR sounds
> suspicious, fwiw

I am not trying to solve all problems, maybe it is wrong, but that
is for someone else to decide.

> 
>>Index: package/libusb/libusb.mk
>>===================================================================
>>--- package/libusb/libusb.mk (revision 19021)
>>+++ package/libusb/libusb.mk (arbetskopia)
>>@@ -36,7 +36,7 @@
>> touch $(LIBUSB_DIR)/.configured
>> 
>> $(STAGING_DIR)/$(LIBUSB_BINARY): $(LIBUSB_DIR)/.configured
>>- $(TARGET_CONFIGURE_OPTS) $(MAKE) CC=$(TARGET_CC) -C $(LIBUSB_DIR)
>>+ $(MAKE) $(TARGET_CONFIGURE_OPTS) -C $(LIBUSB_DIR)
>> $(MAKE) -C $(LIBUSB_DIR) DESTDIR=$(STAGING_DIR) install
> 
> also suspicious DESTDIR
>


Best Regards
Ulf Samuelsson