pkgsrc-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Odd checksum issue on MacOS



On 1/13/23 10:25 AM, Louis Guillaume wrote:
On 1/13/23 3:07 AM, Jonathan Perkin wrote:
* On 2023-01-13 at 02:50 GMT, Louis Guillaume wrote:

On 1/12/23 3:27 PM, Jonathan Perkin wrote:
* On 2023-01-12 at 20:22 GMT, Louis Guillaume wrote:

Seems like TOOLS_PLATFORM.sed=nbsed is generally more appropriate for MacOS. I'll add this to my mk.conf, but maybe it should be the default?

It is, and has been for a long time:

https://github.com/NetBSD/pkgsrc/blob/trunk/bootstrap/bootstrap#L607-L612

The only thing I can think of is that you bootstrapped a very long time ago on a version which happened to not match the previous version of that code.


It hasn't been that long since I re-bootstrapped. Probably 2 years. But as I look at the repo, it shows /usr/bin/sed for TOOLS_PLATFORM:

$ grep sed tools.Darwin.mk
TOOLS_PLATFORM.sed?=            /usr/bin/sed

mk/tools/tools.Darwin.mk will always default to that.  What the bootstrap code does is add e.g.:

   $ grep sed /opt/pkg/etc/mk.conf
   TOOLS_PLATFORM.sed=    /opt/pkg/bin/nbsed

to the mk.conf generated at bootstrap time, and that value is used in preference to the one from tools.Darwin.mk (which uses ?= to mean "only set if not already set).

Somehow it seems that your mk.conf, whether it was not added at bootstrap time or has been removed later, didn't have that line.


I see what you're saying. My mk.conf is def. ancient! I probably should have compared it vs. the bootstrap-generated sample. Would it be more reliable to change the default in the default space, tools.Darwin.mk, though? What would be the reason to maintain the default value with the less effective option only to have an override in mk.conf for everyone?

--
Louis



Hi Jonathan, Roland,

Looks like this came up again today as I did my quarterly builds. I had, in mk.conf, TOOLS_PLATFORM.sed=nbsed.

This caused a rebuild of "perl" to fail. Then after un-doing that libcfg+ failed just the same as it did before.

Seems to be isolated to just libcfg+ (out of the packages I'm using).

Should TOOLS_PLATFORM.sed be set explicitly for this package in the Makefile? Maybe something like...



--- devel/libcfg+/Makefile      30 Sep 2015 20:30:54 -0000      1.11
+++ devel/libcfg+/Makefile      7 Jul 2023 20:18:50 -0000
@@ -15,3 +15,7 @@
 MANCOMPRESSED=         yes

 .include "../../mk/bsd.pkg.mk"
+
+.if ${OPSYS} == "Darwin"
+TOOLS_PLATFORM.sed=nbsed
+.endif



This does work in isolation - maybe there's a better way?

--
Louis


Home | Main Index | Thread Index | Old Index