tbh, if there's a series of (it feels like) 30+
patches doing lots of complex
The CSD patches were numerous and complex, too.
These were merged WITHIN A SINGLE DAY.
The contrast to MSC voice and HNBGW cnpool can hardly be more extreme.
You are asking for a list of patches -- I'd rather not go into details there
even more, but i found it hard to get through code review for the later half of
https://gerrit.osmocom.org/q/cnpool+project:osmo-hnbgw
And the "sensible time stamp" patches
https://gerrit.osmocom.org/q/topic:log_timestamps
https://gerrit.osmocom.org/c/osmo-hnbgw/+/33133 has a
discussion about
parsing of routing area identities and using shared code rather than
https://gerrit.osmocom.org/c/osmo-hnbgw/+/33173 contains a discussion
about naming. Naming *is* important, particularly if it uses terms
which already have well-established meaning, such as (here) what is a
https://gerrit.osmocom.org/c/osmo-hnbgw/+/33178 contains a "maybe add a
log line" spawning a discussion with relatively verbose messages in it.
Of course I agree with all of these points, and it is not like my responses to
the CR were "go away you are stupid". I have very concise reasons for choices
made, and it felt like these were brushed aside. So I explained again...
That might be one of the cases you are referring to?
To be fair it was
a "maybe" comment, so a simple one line response might have been
sufficient.
Reasons for coding choices are usually complex.
So if code review concentrates on inherently marginal items, and me saying so
doesn't help, then we're stuck in a discussion of complex and precise reasons
why a stupendously unimportant line of code was written that way.
All of these discussions are absolutely valid in principle.
But all of these can reach a point of losing touch with what is important.
For example, if my patches *omitted* all logging, no CR would have complained.
What made me write the mail is: a customer is waiting for my merge, and we're
stuck discussing logging. At the same time I am not getting a chance to even
read the pretty complex CSD patches before they were merged.
To me it felt like a double standard being applied, certainly out of
circumstance, not on purpose. Nevertheless I would welcome some agreement that
this wasn't exactly optimal.
So while it was me who merged that series, and I'm
of course sorry if
that created problems, I acted in good faith (see above).
ack
I think if anyone wants to hold back a merge for
whatever reason, make
sure you put a -1 vote in so it becomes impossible to be merged by
yes, but the merge was too quick for even that.
It's not all lost: One can always send follow-up
patches?
Yes, but now it is my work load: i have to modify the status quo and verify
with testing etc (and apply code review), so in practice this is unlikely to
happen. I don't get to drop a comment and continue my own task =)
I always found it mildly fascinating how many options,
choices and
formats (I think primarily you) introduce to the libosmocore logging.
yes, i wish there were one single, sensible format.
We had a lot of options before i started: print category-hex, print category,
print level -- of course everyone wants to see both log level and category, and
obviously the category *name* and not some hex number...
I wanted other things changed because they way logs are structured greatly
affects debugging cycles. For example, if source file and line are printed in
the beginning, it heavily breaks up all "indenting" / lining up of log lines
above and below. Another example is the "extended-timestamp" where it is
extrememly annoying to find the numbers for hour, minute and second: on every
line, i have to count digits from the right. These small invonveniences add up
every minute, every hour, every day of debugging. The only chance to change it
compatibly is to add options.
I have another version of it that does away with the log timezone and the
timestamp format tests, but since that is in my own free time, I'm not quick to
finish it. Probably next time when I am reading logs the next time and find
huge time wasted on the timestamp format, that's when I will feel that it's a
good idea to put effort into getting a useful timestamp...
timestamps, and we are growing more of them. In my
opinion, most of
this is bloat.
For me the old formats are bloat; the only one that makes actual sense is one
showing millis and having separators like 2018-01-16,01:44:34.681
To me, the problem with all those many options is that
it becomes
very difficult to have any software parsing/analyzing osmo* logs
I'm aware of that, and often thought whether we could change the default
logging of osmo to a single accepted format, which has no omissions and is
practical to use.
~N