From: "Bill Page" <bill.page1@sympatico.ca>
To: <gnu-arch-users@gnu.org>
Date: Wed, 1 Dec 2004 00:35:10 -0500
Message-ID: <001a01c4d767$88755250$6501a8c0@Asus>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
Importance: Normal
Subject: [Gnu-arch-users] building arch on cygwin (was: darcs vs tla)
Precedence: list
Sender: gnu-arch-users-bounces+migo=homemail.com@gnu.org
Errors-To: gnu-arch-users-bounces+migo=homemail.com@gnu.org
Status: RO
Content-Length: 3231
Lines: 96
John A Meinel, et al.

Following the recipe below I did successfully build tla on
Cygwin from the listed archives as of 30 November, but in
addition to the unit-unidata test failure below, I am seeing
numerous other test failures:

In hackerlab/tests/fs-tests/unit-file-names.sh
  home...
  expand...

In tla/tests/test-my-id.sh

  Setting my-id: Jane Doe <jane+patches@hacker.gnu.org>
  mkdir: cannot create directory `+': File exists

(At this point I stopped trying to continue running the tests
by commenting out failures.)

In spite of these test failures, tla does seem to be quite
functional in my simple manual tests. Are the test failures
above special cases that do not often occur in practice?
Should I just ignore these failures? Or is there work in
progress to address them?

Also, I notice that the reference to a "binary version available
at http://ct.radiology.uiowa.edu/~jfmeinel/archives"  at 
http://wiki.gnuarch.org/moin.cgi/Native_20WIN32_20Support
is apparently out of date since I no longer find the referenced
files there.  If I recall correctly, I did download an executable
cygwin tla from that location on 19 November. I have been using
it successfully from the last 10 days. Is a new binary now
available somewhere else?

Regards,
Bill Page.

On Thu, 18 Nov 2004 15:02:57 -0600 John A Meinel wrote:

>...
> The easiest way is probably to build from a configuration.
> So you start with:
>
> tla get john at arbash-meinel.com--tla-2004/dists--tla--1.3 tla-1.3
> cd tla-1.3
> tla build-config tla-short-path.cfg
> cd src
> mkdir +build
> cd +build
> ../configure --with-short-paths=1
>     (optionally --prefix=/usr/local or whatever)
> make
> make test
> make install (if you want)
>
> These are the normal steps to build tla, so I didn't repeat them.
> You will also need to have registered the official tla archive,
> since I only modify one portion of it.
>
> ...
>
> I just double checked, and there are 2 caveats with the current system,
> both of them from libhackerlab (which I did not modify.).
>
> 1) The first time you run make it will fail. Just run make again and
> it will succeed. (hackerlab creates a temporary executable to generate
> an output file and then deletes it. But on cygwin/windows the file ends
> with a .exe, so rm cannot find it to delete it. Don't worry, though,
> the dependency is filled)
>
> 2) make test will fail on "unidata-tests". The problem is cygwin picks
> a different line ending if you pipe something through sed, then it does
> if you cat it directly to a file. (sed opens it's input in text mode,
> removing CRLF, while just a redirect preserves the endings).
>
>  The simplest workaround is to edit the file:
>    src/hackerlab/tests/unidata-tests/unit-unidata.sh line 16
>  and change the line from
>   ./unit-unidata > ,tmp2
>  to
>   ./unit-unidata | sed -e '' > ,tmp2
>
> sed -e '' does nothing, but it allows the output to be filtered through
> the same program.
>
> I'm pretty sure with those 2 workarounds, everything plays nice. 


_______________________________________________
Gnu-arch-users mailing list
Gnu-arch-users@gnu.org
http://lists.gnu.org/mailman/listinfo/gnu-arch-users

GNU arch home page:
http://savannah.gnu.org/projects/gnu-arch/