cvs commit: doc/en_US.ISO8859-1/books/handbook/cutting-edge
keramida at ceid.upatras.gr
Thu Dec 23 14:36:11 PST 2004
On 2004-12-23 23:28, "Simon L. Nielsen" <simon at FreeBSD.org> wrote:
>On 2004.12.24 00:14:56 +0200, Giorgos Keramidas wrote:
>>On 2004-12-23 23:06, "Simon L. Nielsen" <simon at FreeBSD.org> wrote:
>>>On 2004.12.23 16:47:15 +0000, Giorgos Keramidas wrote:
>>>> keramida 2004-12-23 16:47:15 UTC
>>>> Describe when it may be necessary to run mergemaster before
>>>> 'make buildworld'. In rare cases, mergemaster may need to be
>>>> run twice: one time before buildworld (to update /etc/group or
>>>> any other files required for a successful buildworld) and one
>>>> time after installworld (to merge all the src/etc updates to the
>>>> /etc directory of the system).
>>>> Point to /usr/src/UPDATING for details of when and why this is
>>> Are you really sure that is correct? AFAIK it is only necessary
>>> before installworld, which is also what UPDATING says.
>> Yes, I am.
>> I asked Doug about this, and he explained to me that there are rare
>> cases when mergemaster -p has to run *before* buildworld too (i.e. when
>> the buildworld stage depends on the existence of a specific user group).
>> These cases are indeed very _rare_. I haven't bumped on anything of the
>> sort for a long time now (not once since I updated my 5.2-CURRENT system
>> to HEAD some time in May 2004, for instance). If UPDATING says that
>> mergemaster -p is necessary though, it definitely *is*. This is why I
>> also pointed at UPDATING as the authoritative reference :-)
> Fair enough, but the examples to build/install FreeBSD in UPDATING
> does not have mergemaster -p before buildworld, so I guess those
> should be updated too ?
Probably yes, but I am not sure about his. The process described now in
UPDATING is ok. It's only for _some_ cases that an extra mergemaster
run might be necessary. If these are described near the head of
UPDATING, is it really necessary to duplicate the same information at
More information about the cvs-doc