[PATCH] tests: fix test-git-import.out

[Giorgos Keramidas]

On 2007-07-15 15:12, Will Maier <willmaier at ml1.net> wrote:
>On Sun, Jul 15, 2007 at 10:42:35PM +0300, Giorgos Keramidas wrote:
>> IIRC, the patch(1) utility on OpenBSD is not based on GNU
>> patch(1), but it is a BSD-licensed rewrite.
> 
> OpenBSD actually uses Larry Wall's original patch(1), though a few
> modifications have been made in the time it's been in the tree. GNU
> patch(1) started from this code, but has since diverged.

Ah, I see now.  Thanks for the helpful history fill-in :-)

>> Of course, this is irrelevant, but if it's easy to support the new
>> <TAB>-based patch style in FreeBSD it's a worthy goal.  It will be
>> nice to be able to support Mercurial with "standard base system"
>> tools.
> 
> That addresses the specific issue, but not the larger problem of
> relying on system tools that are often inconsistent. Almost a dozen
> tests fail on OpenBSD right now because of the system's patch
> implementation. Most of these failures are cosmetic (meaning the
> operation succeeds but the output differs), but I'm still looking into
> a few that show different/broken functionality.

Only the TAB-dependent git-import test fails on FreeBSD and a few others
(mostly related to different errno descriptions) on Solaris.  While I
agree that it would be nice to have a portable set of "tools" around
Mercurial, I am also ok with having a fairly "standard", common set of
very basic tools like diff/patch/merge and then allow building extra
functionality on top or instead of that.  That's probably because I am
spoiled by the "base system" of the BSDs :)

The cron job which I run locally on a Solaris 10 installation to clone a
"crew" mirror and run the tests reports on Solaris today:

%%%
keramida at fingon:/home/keramida/cron.d$ sh mercurial-test.sh
2007-07-16 12:24:50 -- Gate repository is at /home/keramida/hg/mercurial/crew
2007-07-16 12:24:50 -- Creating temporary test workspace at /tmp
2007-07-16 12:24:50 -- Initializing test workspace at /tmp/hgtest-aG9B
2007-07-16 12:24:50 -- Pulling changesets from gate repo at /home/keramida/hg/mercurial/crew
pulling from /home/keramida/hg/mercurial/crew
requesting all changes
adding changesets
adding manifests
adding file changes
added 4886 changesets with 9279 changes to 662 files
(run 'hg update' to get a working copy)
2007-07-16 12:25:02 -- Gate tip is d77accdd236e52c21496b843acda0091d099780a
2007-07-16 12:25:02 -- Checking out changeset d77accdd236e52c21496b843acda0091d099780a
580 files updated, 0 files merged, 0 files removed, 0 files unresolved
2007-07-16 12:25:02 -- Running the Mercurial test suite.
2007-07-16 12:25:02 -- Extra run-tests.py flags: (none)
.............
ERROR: test-bad-pull output changed
--- Expected output
+++ Test output
@@ -1,4 +1,4 @@
-abort: error: Connection refused
+abort: error: Network is unreachable
 255
 copy: No such file or directory
 abort: HTTP Error 404
...................................................
ERROR: test-git-import output changed
--- Expected output
+++ Test output
@@ -62,4 +62,3 @@
 a874b471193996e7cb034bb301cac7bdaf3e3f46 644 mbinary2
 % filenames with spaces
 applying patch from stdin
-foo
...........
ERROR: test-http-proxy output changed
--- Expected output
+++ Test output
@@ -43,4 +43,4 @@
 added 1 changesets with 1 changes to 1 files
 1 files updated, 0 files merged, 0 files removed, 0 files unresolved
 %% bad host:port for proxy
-abort: error: Connection refused
+abort: error: Network is unreachable
...................................................................................
ERROR: test-static-http output changed
--- Expected output
+++ Test output
@@ -1,4 +1,4 @@
-abort: Connection refused
+abort: Network is unreachable
 255
 copy: No such file or directory
 changeset:   0:53e17d176ae6
................
# Ran 174 tests, 0 skipped, 4 failed.
2007-07-16 12:30:08 -- Removing test workspace from /tmp/hgtest-aG9B
keramida at fingon:/home/keramida/cron.d$
%%%

I can live with the "Network is unreachable" failures, but the
test-git-import part is kind of annoying.  Since it's only _one_ test
that's failing, it would be nice to be able to fix just this one instead
of reinvent the entire "world" of the GNU diffutils :-)

- Giorgos