Comments for Squirrel's

Comment on Careful with that PrintWriter by Christoffer Hammarström

Christoffer Hammarström — Sat, 10 Dec 2011 17:09:12 +0000

You’re using the wrong class. To bridge from Writer to OutputStream, you use an OutputStreamWriter *with*the*correct*encoding*specified*, not a PrintStream.

Comment on TDD on Existing Code by Tomek

Tomek — Fri, 02 Dec 2011 13:11:51 +0000

Konrad, what you suggest to do is to add tests to existing codebase. This is fine, valuable, highly recommended etc. etc., however (IMHO), this is not TDD.

Your “definition” of TDD (It is introducing small, fast, focused (as much as possible…), automated tests that become “live” specification and documentation, and protect you from regressions.) is all about automated tests. I disagree – TDD is about driving your design. You do not design existing code, you only add tests to it. Which is fine, but not TDD.

Yeah, I’m a zealot, I know.

–
Cheers,
Tomek Kaczanowski

Comment on Hibernate Cache Is Fundamentally Broken by Konrad Garus

Konrad Garus — Fri, 02 Sep 2011 19:00:17 +0000

Marcos: You’re right. Indeed, I dived so deep that I think I could create a patch. I did not mostly because I see a few possible resolutions:

1. “Won’t fix” – Hibernate team can decide it’s not a bug, or is not important enough, and you have to take full responsibility for this kind of cache “poisoning”.

2. Treat entities with different schema as misses, and fetch from DB:
2a. Use serialVersionUID
2b. Use a new class-level attribute (@MappingVersion, )
2c. Use both: 2b is preferred, but if not present use 2a.

While I think I am capable of implementing it myself, I don’t think such an important and disputable change in core can be done by a lone developer who’s never contributed and may not see the whole big picture.

Comment on Hibernate Cache Is Fundamentally Broken by Marcos Eliziario Santos

Marcos Eliziario Santos — Fri, 02 Sep 2011 18:13:25 +0000

Good thing, but the time you’ve spent diagnosing this problem and writing this, could have been spent writing a patch. That’s how Open Source is supposed to work.

Comment on “Processing in Clojure” Made Functional by Niels Egberts

Niels Egberts — Wed, 22 Jun 2011 18:59:46 +0000

I would like to add that if you want to let the whole application exit when the user closes the window you can add the following line to the last doto.

(.setDefaultCloseOperation (javax.swing.JFrame/EXIT_ON_CLOSE))

Comment on Do Leave Failing Tests by Ray

Ray — Tue, 07 Jun 2011 03:43:07 +0000

Ray, I do the same thing. I review the commit diff, to understand what I did.

Comment on Do Leave Failing Tests by Alex Skorulis

Alex Skorulis — Mon, 06 Jun 2011 03:36:48 +0000

I started doing this a while ago. Lately I’ve been putting what I have to do next straight into code without commenting it. Come monday morning the compile error will take me straight there and the message should be enough to get my brain ticking along.

Comment on Do Leave Failing Tests by Ray

Ray — Sat, 04 Jun 2011 13:08:27 +0000

Deadly post – I’ve gotten similar value out of leaving uncomitted code, so when I sit down, I just view the diff, and it’s as if the stack trace from the work I was doing suddenly loads straight back into my head.

Comment on Do Leave Failing Tests by Geek

Geek — Sat, 04 Jun 2011 11:56:37 +0000

What I do is, I just stop writing code and start writing text explaining what I was doing when I left off.

I use the
#error
macro to ensure that it can’t compile, and I leave a little brain dump of my state at the point where I was working.

It is better to leave it in a state that won’t compile (and not checked in, of course so you don’t break the build) than to leave something “working” that isn’t really complete and thus has subtle bugs… or to write crap code just to “finish” it in a rush or when you’re tired.

Comment on Do Leave Failing Tests by John M

John M — Sat, 04 Jun 2011 10:28:22 +0000

@foobaz

If I have to fix a bug in released code – I will go to code that is released (usually I have a clean copy of that on computer anyway, but if need be I’d check it out) – as I would intend releasing just that fix and just that fix as that would be safest, require least testing and quickest to release.

If I am working on new features/release then I would expect that to go through a suite of regression and user acceptance tests irrespective of what my automated tests show.

So it is irrelevant what state I leave the code I am working on just to worry about some hypothetical very low probability issue is