On Saturday-201404-12, 1:13, Karl Williamson wrote:
blead contains many fixes for EBCDIC support that aren't in 5.16, such
as to charclass_invlist.h. There is no sense in you trying to refix
these. The branch khw/ebcdic has even more fixes, mainly to tests to
make them EBCDIC aware. Last we tried it we were up to about 92% of
core tests passing, most of the failures were due to one bug, I believe,
involving pack/unpack, and to-be-expected porting tests that just need
to be changed to be skipped on non-ASCII platforms.
Sounds good!
Karl: to kick things off, maybe you could prepare a snapshot.tar.gz of
your branch for Yaroslav, something concrete to try out immediately? I
have no idea (but I have my doubts) that the z/OS UNIX tools might not
include git. This would complicate things somewhat.
Yaroslav: take a look at http://www.test-smoke.org/ and
http://www.test-smoke.org/FAQ.html - that is what you need to set up an
automated periodic "smoke build". A continuous build would be the best,
daily build would be okay, a weekly build kind of acceptable. (Of
course depends on how powerful systems you have: on older systems it was
not unusual for a single build to take more than a day!)
This is the branch that ongoing work should start with. I am willing to
work with you to get it to really work, but only if we identify a smoker
for ongoing testing. Without this, the pumpking has decreed, and I
fully agree, EBCDIC support will be dropped. Instead of now working to
root out the bugs you've found, which are likely already fixed in blead,
we should be working to get a smoker identified, one with a long-term
commitment to its continued availability.
