2018. 09. 03. 10:53 keltezéssel, Vlad Khorsun írta:
Issue (a) is fixed now (see above) but issue (b) is hard to fix and i
need some time
to find if it can be fixed and how to do it. Probably, it should not be
fixed
at all (hi, Adriano ;)). I going to try this test with execute
statement, just to
see what happens.
Please fix it. If not, the WAIT feature is useless. Or not?
PS In ReadConsistency = 0 mode test case produced a lot of backversions
of single record
that was updated many times. It happens because "main" transaction
blocks garbage
collection.
In ReadConsistency = 1 (after i added retry on update conflict handling
at P-SQL level)
it produces just 2 backversions - thanks to the intermediate garbage
collection.
"2" backversions better than "lot of". But only if the WAIT feature
working. Default value of ReadConsistency is illogical for me. The
documentation say "To help test existing applications...". If I want to
test a new feature I turn it on. Currently turn off to eliminate "update
conflicts...". (My own test application execute 1 statement/second which
is not a high load and got "update conflicts..." instantly if execute
two instances in parallel from it.
A defaulted non properly working new feature is not a good message (for
me). Develop it, test it and if everything works with it as before, set
as default. But I am just a user. :D
Gabor
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
