Re: ampere2's main-armv7 crashed after only 3 min 51 sec, nothing started to build; libdm.so.{6->7} issues

On Oct 13, 2024, at 13:18, Ronald Klop <[email protected]> wrote:

> On ampere3 armv7 builds crashed in the same way. 
> 
> https://pkg-status.freebsd.org/ampere3/
> 
> Regards,
> Ronald
> 
>> Van: Mark Millard <[email protected]>
>> Datum: 13 oktober 2024 22:13
>> Aan: Antoine Brodin <[email protected]>, FreeBSD Mailing List <[email protected]>, FreeBSD ARM List <[email protected]>
>> CC: Philip Paeps <[email protected]>
>> Onderwerp: ampere2's main-armv7 crashed after only 3 min 51 sec, nothing started to build; libdm.so.{6->7} issues
>> 
>> main-armv7 crashed before it was at a stage that has public log files to look
>> at.
>> 
>> I report this mostly because the currently distributed main-packages for ports
>> that involve use of libmd.so.6 are broken because main now has/uses libmd.so.7
>> instead --so libmd.so.6 tends to be missing. Even when libmd.so.6 and
>> libmd.so.7 both exist, the pkg (non -static) command is broken from the
>> recursive dependencies ending up referencing both. (pkg-static works but pkg
>> does not.)
>> 
>> It might be that the crash is because of the libmd.so.6 to libmd.so.7 change
>> in main. But I've no access to logs to look at.
>> 
>> It would be good to avoid having another main-arm64 bulk -a happen before
>> main-armv7 has a chance to produce main-armv7 packages with libmd.so.7
>> references so that ports are again good for (modern) main-FreeBSD [so: 15].
>> 
>> But it seems that, for main-arm* port-packages, either:
>> 
>> 0) various distributed ports are incompatible with libmd.so.7 based
>>     main-FreeBSD (the current context for main-armv7)
>> 
>> vs.
>> 
>> 1) various distributed ports will be incompatible with libmd.so.6
>>      based main-FreeBSD (so: older main/15 FreeBSD installations):
>>      the future contexts for main-armv7 and main-arm64 (and more).
>> 
>> 
>> FYI: The libmd.so.{6->7} change dates back to 2024-Sep-30 in FreeBSD's main.
>> But the __FreeBSD_version 150002{3->4} change did not happen until
>> 2024-Oct-02.
>> 
>> 
>> main-powerpc-default's status for such:
>> 
>> Looks like FreeBSD:15:powerpc/latest/ was last updated on 2024-Mar-07
>> and so might not have the issue --by no longer having port-package
>> updates of any kind. (32-bit powerpc is not to be supported by
>> releng/15.0 as I understand: The only 32-bit platform will be armv7
>> as I understand.)

Looks like I should have done a wider exploration instead
of presuming just main-armv7 was broken.

An interesting point is that none of the *-armv7-quarterly
jails have failed so far. Just the *-armv7-default have
failed:

main-armv7-default p149fe86b8e79_s149e1af6a ,
141releng-armv7-default 1f84c1fae602 ,
133releng-armv7-default 1f84c1fae602 .

I'll note that the arm64 ones did not fail:

main-arm64-default p149fe86b8e79_s149e1af6ae4
141arm64-default 1f84c1fae602 ,
133arm64-default 1f84c1fae602 .

133releng-armv7-default is the oldest failure for this issue:
Sun, 06 Oct 2024 23:33:31 GMT start with elapsed: 00:05:26 .

133releng-armv7-default had a  prior build: Thu, 19 Sep 2024 19:55:19 GMT 58:03:54

141releng-armv7-default had no prior build attempts ( was: 140releng-armv7-default ).

     main-armv7-default had a  prior build: Thu, 03 Oct 2024 04:21:42 GMT 61:50:05


I wonder what a common change is across the failing examples.


===
Mark Millard
marklmi at yahoo.com