Re: [9fans] mk all

Re: [9fans] mk all

[forsyth]

> 8^l  -o 8.nntpfs nntpfs.8
> nntpresponse: undefined: strecpy in nntpresponse
>
> mk: 8^l  -o ...  : exit status=rc 7399:8l 7401:error

# try:
cd /sys/src/libc
mk install

Re: [9fans] mk all

[Lucio De Re]

On Sun, Apr 29, 2001 at 11:45:28AM +0100, forsyth@caldo.demon.co.uk wrote:
>
> # try:
> cd /sys/src/libc
> mk install

I think a full rebuild is possible, (so far) without hitches, based on

	bind -c /n/other/sys/src /sys/src
	bind    /n/other/sys/include /sys/include
	bind    /n/other/386/include /386/include
	bind -c /n/other/386/lib /386/lib
	bind -c /n/other/386/bin /386/bin
	cd /sys/src
	mk install

The only weird event was a "memcpy" warning, very early in the
creation/update of /386/lib/libc.a - complained of a duplication (I
don't have the error message) and claimed it did not build the archive
(again, from memory).  I saw it twice, but not after a recent

	mk clean

Sadly, mk all hardly makes "all", leaving a lot of work for "mk
install".  I suspect GS will eventually flatten the cpu server I'm
running the build on, but that's not too critical, except that from
Cape Town I'm not going to be able to recover if it crashes in any
serious fashion in Johannesburg :-(

GS is what's being compiled at this moment :-)

++L

[9fans] mk all

[Lucio De Re]

I figured I had the wrong header files, eventually by

	bind /n/other/sys/include /sys/include
	bind /n/other/386/include /386/include

I managed to get past my first compilation error(s).  I must
confessed that I'm not convinced I have the same MD5 sums as Dave,
but that will ride...

Now I get:

> 8^l  -o 8.nntpfs nntpfs.8
> nntpresponse: undefined: strecpy in nntpresponse
>
> mk: 8^l  -o ...  : exit status=rc 7399:8l 7401:error

after much other stuff (copied, not cut-and-pasted, but it seems
accurate).

Also, I sure wish the "mk all" operation installed the libraries
in a staging area and let "mk install" actually move them to their
final location.  This would also require the mkfiles to instruct
the loader to search the new location, but that should not be too
great an issue.

I should imagine one can create a staging area:

	bind -c /n/other/386/lib /386/lib

and perhaps the /sys/src mkfile could arrange all of the above
namespace adjustments if its working directory is of the form
<prefix>/sys/src.

Any pointers to solving the small hiccup above are welcome.  This
is after installing the latest full release and applying the
03270425a.9gz update (there are lots of date inconsistencies I
chose to override, but when I didn't, the error was the same).

++L