pkgsrc-Users archive

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

Re: pkg_rolling-replace and libtool fallout



Hauke Fath <hauke%Espresso.Rhein-Neckar.DE@localhost> writes:

> On Mon, 3 Aug 2020 12:06:22 +0200, Hauke Fath wrote:
>> during an ongoing pkg_rr run, I am getting a lot (as in: every fourth 
>> or fifth package)  of build breaks like
>> 
>> [...]
>> libtool: compile: unable to infer tagged configuration
>> libtool:   error: specify a tag with '--tag'
>
> The important missing bit: Manually building and replacing the packages 
> in question always succeeds.

That is bizarre.  If you read pkg_rr, it's a shell script which
basically loops doing:

  a lot of set operations, written in pretty sick sh, to figure out what
  to replace and in what order

  in the chosen directory, does "make package replace clean"

I have never had a report of this kind of difference before.  reports
have been one or the other of:

  pkg_rr's set calculations do $RANDOM_WRONG_THING

  some package doesn't build

as to your invocation:

  nice -n ${NICENESS} pkg_rolling-replace -rusv ${_excludes} 2>&1 | tee 
  ${_buildlog}

  from a shell script in tmux. Some packages (prominently firefox, 
  thunderbird) appear to object to that (maybe the redirect?), with 
  usually obscure errors, and insist on a manual build, which then 
  succeeds. But the current libtool fallout is unusual.

have you then done

  nice -n ${NICENESS} make package replace clean 2>&1 | tee ${_buildlog}

in exactly the same way?

I do notice that some packages try to read from stdin, and get hung up
if they are in bg and stdin is the tty, so I redirect stdin from
/dev/null.

FWIW, I can build firefox (well, when it builds) with pkg_rr just the
same as manual.  But I redirect stdin.


Home | Main Index | Thread Index | Old Index