[bug-inetutils] The quirks of the bootstrap script

[Mats Erik Andersson]

Hello there,

I could use a description of the limitation inherent in
the script "bootstrap", as it applies to GNU Inetutils.
I do understand that it comed from GNUlib, but I do also
observe some effects that trouble me.


The thing is that there are no sufficiently recent
autotools available in OpenBSD. Thus my work in porting
the present source tress imposes a need to perform
the bootstrapping with GNU/Linux, followed by some
transfer to the BSD system.

My intuitive expectation, called for by the word "bootstrap",
is that the action should generate a resulting content
equivalent to an official release archive, although littered
by various VCS files, in this case for the book keeping of Git's.


After bootstrapping, I should be able to use a transfer command

    $ rsync -Ca inetutils  address@hidden:~/gnu/build/

Under the present circumstance this does not work. The source
can be configured (my standard choice is the bare modifier
CFLAGS="-Wall"), but compilation is always aborted since

    lib/unlocked-io.h

is dangling soft link, lacking a target. I interpret this
as a misconceived bootstraping script, until better educated!


If I instead use a waistful transfer

    bsdbox$ nc -l 1234 | tar -xf -

    debian$ tar -cf - inetutils/ | nc -x bsdbox 1234

then I do get a source tree where compilation on OpenBSD can
indeed be completed, given the two changes I have reported on earlier.


When performing the bootstrapping under GNU/Linux, there are
many messages passing by, but I have in particular observed
some warnings that certain header files are not maintained
using Git, while others are. I also notice that the particular
file "lib/unlocked-io.h" ought to be related to the fact that
"bootstrap.conf" mentions "unlocked-io" as one of the modules
from GNUlib on which GNU Inetutils itself depends.


It might be the case that the whole matter is due to a
misconception on my side, but I would like to get the
project members to reflect on this matter, in order that
there be no implicit assumption buried in the code
or in the infrastructure.


Best regards,

Mats E A