svn commit: r413727 - in head: Mk/Uses cad/irsim cad/magic cad/netgen cad/netgen/files cad/xcircuit lang/expect lang/itcl lang/itcl4 lang/itcl4/files lang/nx www/mod_rivet www/tclhttpd www/tdom www...

John Marino freebsd.contact at marino.st
Sat Apr 30 16:20:35 UTC 2016


On 4/30/2016 4:06 PM, Pietro Cerutti wrote:
> On 2016-Apr-30, 12:32, John Marino wrote:
>> On 4/21/2016 1:33 PM, Pietro Cerutti wrote:
>>> Author: gahr
>>> Date: Thu Apr 21 11:33:16 2016
>>> New Revision: 413727
>>> URL: https://svnweb.freebsd.org/changeset/ports/413727
>>>
>>> Log:
>>>     Support Tk in USES=tk:tea. Convert more ports + minor fixes and modernization.
>
> [snip]
>
>> [snip]
>>
>>>
>>> Modified: head/lang/expect/Makefile
>>> ==============================================================================
>>> --- head/lang/expect/Makefile	Thu Apr 21 11:18:13 2016	(r413726)
>>> +++ head/lang/expect/Makefile	Thu Apr 21 11:33:16 2016	(r413727)
>>> @@ -14,25 +14,16 @@ LIBEXPECT_VER=	${PORTVERSION:S/.//g}
>>>
>>>    WRKSRC=		${WRKDIR}/${DISTNAME}
>>>
>>> -USES=		tcl autoreconf
>>> +USES=		tcl:tea autoreconf
>>>    USE_LDCONFIG=	yes
>>> -GNU_CONFIGURE=	yes
>>> -CONFIGURE_ARGS=	--enable-shared \
>>> -		--exec-prefix=${PREFIX} \
>>> -		--datadir=${DATADIR} \
>>> -		--with-tclconfig=${TCL_LIBDIR} \
>>> -		--with-tclinclude=${TCL_INCLUDEDIR}
>>> -CONFIGURE_ENV=	ac_cv_sys_long_file_names=yes
>>> +CONFIGURE_ARGS=	--datadir=${DATADIR}
>>>    PLIST_SUB+=	EXPECT_VER=${PORTVERSION} \
>>>    		LIBEXPECT_VER=${LIBEXPECT_VER}
>>
>> [snip]
>>
>> This line, "CONFIGURE_ENV= ac_cv_sys_long_file_names=yes" is unrelated
>> to tea modification and it should not have been removed.
>
> Right, it was related to the "minor fixes and modernization" part. My
> tests - poudriere runs on 9 and 10, i386 and amd64 - didn't suggest this
> was needed. If you have evidence of the contrary, please share your
> findings.

Poudriere apparently can't detect the file system violation because 
according to Matt poudriere builds with non-root by default and the user 
doesn't have the ability to write to /usr/local/lib even though it's 
attempted.

you can detect it with "synth test" though.  The configure script writes 
to /usr/local/lib as part of the conftest to determine if the file 
system supports long names.  This conftest is in violation -- there 
should be no writing to localbase during the configure and build phases.

John






More information about the svn-ports-head mailing list