On 2006-11-07, at 14:48 EST, Benjamin Shine wrote:
On Nov 7, 2006, at 7:50 AM, P T Withington wrote:
Comments:
1) Your editor apparently sticks lots of attributes into text
files, and your tar bundles them up as ._<filename>. Just an
annoyance, I guess, but it would be nice to teach your tar to not
bother with these attributes for the purposes of review. (What
tar are you using?)
I'm using GNU tar in /usr/bin/tar... educating it would require
changing svn-bash.sh. I'll try a change locally, and see if it
fixes this without breaking other things. (I'm concerned about the
chance of mucking things up when adding exclusions that start with
the potentially regexp-meaningful "._")
I'm using the same, so something else is going on. I'm not sure that
filtering for ._ will have any effect; I am suspicious that this is a
fictitious file name that represents the second fork of a Mac file.
It probably gets written for any file that has more than one fork.
We may not have seen it before because the editors we use don't write
any meta information into text files....
2) Rather than output discrete information about the build, etc.
for each log entry, why not just once per run output
canvas.versionInfoString()?
It was easier to write the parsing code with the output tagged and
tokenized with spaces and equal signs.
I put all the log info into each log entry so I can have a
stateless log analyzer. I can let the log file accumulate across
different runs, and easily select log entries that interest me,
with a variety of per-line selection tools.
I've been opting for the easiest-to-program method, allowing some
egregious space and time hits in order to get to a minimal working
version fastest. Your rules for optimization (1. Don't do it. 2.
[for experts only] Don't do it, yet.) have loomed large in my mind,
along with Jim and Henry's, "what's the simplest thing that would
possibly work?"
A reasonable strategy.
Thanks for the comments.
Otherwise, approved.
On 2006-11-06, at 18:49 EST, Benjamin Shine wrote:
Change 20061106-ben-p by [EMAIL PROTECTED] on 2006-11-06 15:33:58 PST
in /Users/ben/src/svn/openlaszlo/branches/legals/test
Summary: Extensions to performance logger to record more info
about the runtime
New Features:
Bugs Fixed:
Technical Reviewer: ptw (pending)
QA Reviewer: (pending)
Doc Reviewer: (pending)
Documentation:
Now we report tons of information to be logged by the
reporter.jsp, including
all the numbers about a particular meter (with number of runs,
min, max, and
uncertainty). We also record the User-Agent http header, the ip
address of the
client, and the runtime.
Log info now looks like this:
testname=newView suitename=viewperf duration=1.5699999999999998
count=10 min=0.95 max=2.6 uncertainty=0.5390732788777421
runtime=dhtml buildid=2405 debug=false
builddate=2006-11-06T13:49:41-0800 browserinfo=Mozilla%2F5.0+%
28Macintosh%3B+U%3B+Intel+Mac+OS+X%3B+en-US%3B+rv%3A1.8.1%29+Gecko
%2F20061010+Firefox%2F2.0 comment=127.0.0.1
Dialed down number of runs per test to make the tests complete
faster.
Release Notes:
Details:
Tests:
http://localhost:8080/legals/test/lfc/perf/viewperf.lzx?lzr=dhtml
http://localhost:8080/legals/test/lfc/perf/viewperf.lzx?lzr=swf7
http://localhost:8080/legals/test/lfc/perf/viewperf.lzx?lzr=swf8
Examine the file in test/lfc/perf/results.txt
Files:
M lfc/perf/cssperf.lzx
M lfc/perf/typeof.lzx
M lfc/perf/viewperf.lzx
M lfc/perf/reporter.jsp
M lfc/perf/timing-classes.lzx
M lfc/perf/reporter.lzx
Changeset: http://svn.openlaszlo.org/openlaszlo/patches/20061106-
ben-p.tar
Benjamin Shine
Software Engineer, Open Laszlo / Laszlo Systems
[EMAIL PROTECTED]
Benjamin Shine
Software Engineer, Open Laszlo / Laszlo Systems
[EMAIL PROTECTED]
