On 1/6/2016 11:47 AM, Joe Mistachkin wrote:
It appears that all the remaining failures, except "merge_multi-4" and
"merge_renames-5" are caused by not having an open checkout available
(i.e. the "th1.test" and "file1.test" test files must be run from within
the Fossil checkout).
If the test framework is supposed to work (as I believe it should)
outside of an open checkout of the fossil repository itself, then these
cases need to be improved to supply a suitable open checkout for context.
I personally don't like the idea of a test suite that makes changes to
the true project repository, and the comment near "make test" in the
main.mk *really* doesn't inspire confidence:
> # WARNING. DANGER. Running the test suite modifies the repository the
> # build is done from, i.e. the checkout belongs to. Do not sync/push
> # the repository after running the tests.
> test: $(OBJDIR) $(APPNAME)
> $(TCLSH) $(SRCDIR)/../test/tester.tcl $(APPNAME)
IMHO, it would be well worth the effort to invalidate and remove that
comment. If required, it could be replaced by advice to use "make test"
only for out-of-tree builds.
Because of that comment and Richard's advice to test out of tree to
avoid that issue, I haven't run a test in tree. I suppose I should pull
a fresh clone and try it to better understand the issue.
I'm not sure about "merge-utf-27-23" and "merge-utf-27-32". They failed
here last night as well and I'm not sure why.
I looked at those last night after grep failed to quickly tell me that
they are from merge2.test, and it feels like this is pointing at a bug,
not a false positive. I'll dig a little more later.
--
Ross Berteig r...@cheshireeng.com
Cheshire Engineering Corp. http://www.CheshireEng.com/
+1 626 303 1602
_______________________________________________
fossil-dev mailing list
fossil-dev@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/fossil-dev
