RFR (s): 8026849 - Fix typos in the GC code, part 2
kirk at kodewerk.com
Mon Feb 3 18:59:28 UTC 2014
The format in question hasn’t changed in a long time, in fact ever! Why do we suddenly need to change it now? The change doesn’t relate to new functionality nor does it represent an improvement in the quality of information in the logs nor does it fix functionality (ie. fixes a bug and calling this a bug is hyperbole IMO). I would never complain about any change that meets this standard. FWIW, I’ve been looking at *all* of the rules that I use and it’s amazing that in some cases I’m looking at more than a half a dozen fundamentally different formats for expressing the *exact same information*, no improvement in the quality of information nor does it expose new functionality. And that’s not including taking into account all the spelling mistakes that have cropped up, missing carriage returns, records corrupting each other. You have to admit that CMS logging is a disaster! So all I’m asking is that before changes are made in the log format please lets first ask; is this change really improve the quality or am I pandering to personal preferences. You prefer a ‘,’, I like the ‘ ‘.
My suggestion quite some time ago was to come up with a consistent and easily parseable format. I’ve been willing to dedicate some time to do so but it was suggested that I wait. G1 is some what better but sadly, depending upon which flags you set you get records crashing into each other (RSet diagnostics as an example). My first suggestion is lets fix that problem first.
On Feb 3, 2014, at 5:46 PM, Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
> Hi Kirk,
> On Thu, 2014-01-30 at 14:44 +0100, Kirk Pepperdine wrote:
>> Hi Jesper,
>> Understood. Did this people express their views in a forum where we all can see
>> and understand why this comma is so important to them?
> here are some thoughts about this issue. There is no particular big
> discussion about this, mostly I think Jesper's closing of the bug as
> Won't fix was too hasty.
> My main gripe is that this issue is a bug. Whatever it is, it should be
> fixed at the best possible time. It's an ugly wart.
> The other consideration is that the impact of the change is within my
> expectations of the stability of the Hotspot log.
> I.e. in my view there are only modest expectations on keeping the GC log
> unchanged for a long time: it reflects the internal workings of the GC
> at the current time, which requires some breakage now and then.
> So I do not feel that this particular change is outside the limits of
> this regular expected breakage. See the examples about adding or
> changing the contents of phases. I agree that those changes would give
> you additional information (i.e. personal benefits), and your programs
> in particular do not care about typos in the logs, but it is not only
> your programs that look at them.
> We also need to work with old versions of log files. ;)
> If you find other similar issues, or have particular ideas etc., please
> post them here, file bug reports, and (best) accompany them with fixes.
> As for the timing of this change: currently one release is on the way
> out (8), and we are starting to push changes for the next (8u20, 9).
> This should actually be optimal, as it gives the users the most time to
> react to such changes.
> When else do you expect us to fix such "trivial" stuff?
> Afair, we also asked Jesper at the end of the 8 release cycle to keep
> off from pushing all these typo changes.
More information about the hotspot-gc-dev