matplotlib 64 bit installs never work

a clue as to why:
functional ft2font.so:

$ otool -L /Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/site-packages/matplotlib/ft2font.so
/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/site-packages/matplotlib/ft2font.so:
/usr/local/lib/libfreetype.6.dylib (compatibility version 10.0.0, current version 10.22.0)
/usr/lib/libz.1.dylib (compatibility version 1.0.0, current version 1.2.3)
/usr/local/lib/libstdc++.6.dylib (compatibility version 7.0.0, current version 7.14.0)
/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 125.2.11)


nonfunctional ft2font.so:

$ otool -L /Users/adam/repos/yt/yt-i386/lib/python2.7/site-packages/matplotlib/ft2font.so
/Users/adam/repos/yt/yt-i386/lib/python2.7/site-packages/matplotlib/ft2font.so:
/Users/adam/repos/yt/yt-i386/lib/libfreetype.6.dylib (compatibility version 13.0.0, current version 13.2.0)
/Users/adam/repos/yt/yt-i386//lib/libz.1.dylib (compatibility version 1.0.0, current version 1.2.3)
/usr/local/lib/libstdc++.6.dylib (compatibility version 7.0.0, current version 7.14.0)
/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 125.2.11)
/usr/local/lib/libgcc_s.1.dylib (compatibility version 1.0.0, current version 1.0.0)


The culprit is the difference between those two (maybe?):

$ file /usr/local/lib/libgcc_s.1.dylib
/usr/local/lib/libgcc_s.1.dylib: Mach-O universal binary with 4 architectures
/usr/local/lib/libgcc_s.1.dylib (for architecture i386): Mach-O dynamically linked shared library i386
/usr/local/lib/libgcc_s.1.dylib (for architecture x86_64): Mach-O 64-bit dynamically linked shared library x86_64
/usr/local/lib/libgcc_s.1.dylib (for architecture ppc): Mach-O dynamically linked shared library ppc
/usr/local/lib/libgcc_s.1.dylib (for architecture ppc64): Mach-O 64-bit dynamically linked shared library ppc64
$ otool -L /usr/local/lib/libgcc_s.1.dylib
/usr/local/lib/libgcc_s.1.dylib:
/usr/local/lib/libgcc_s.1.dylib (compatibility version 1.0.0, current version 1.0.0)
/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 88.3.9)


Major mac problems

The errors are, in short:

• Browser stops responding / starts returning "page not found"
  (indicating a failure of mDNSResponder)
  
• killing mDNSResponder sometimes brings browser back, but more often
  leads to a partial system freeze (some windows don't respond, can't
  switch between windows except by clicking)
  
• /var/log/system.log gets flooded with "too many files open" errors.
  
• somewhere in here the Dock fails
  
• killing Google Chrome and/or the Dock fails; the process never halts
  (even kill -9 + kill -s SIGCHLD)
  
• usually one or two crash reports pop up, at least one of which is for
  crash_reporter
  
• system.log stops getting flooded, but the browser and Dock never recover
  

The only message in system.log that gives me any hint about what might be happening is occasionally a freeze was resolved at the same time as this message: Jun 22 19:02:09 eta Quicksilver[93771]: Multiple Scans Attempted but it doesn't seem to change the situation if quicksilver is open or not. Google has nothing on this issue, either, except for the quicksilver source code, so evidently it has not caused problems for other people system.log was also being flooded with this message:

Jun 23 08:57:19 eta postfix/master[99954]: fatal: open /dev/null: Bad file descriptor
Jun 23 08:57:20 eta com.apple.launchd[1] (org.postfix.master[99954]): Exited with exit code: 1
Jun 23 08:57:20 eta com.apple.launchd[1] (org.postfix.master): Throttling respawn: Will start in 9 seconds

so I disabled my postman:

Jun 23 08:57:27 eta sudo[99955]: adam : TTY=ttys006 ; PWD=/Users/adam/proposals/alma ; USER=root ; COMMAND=/bin/launchctl unload -w /System/Library/LaunchDaemons/org.postfix.master.plist

These errors:

Jun 23 09:06:40 eta Dock[99877]: kCGErrorIllegalArgument: CGSSetWindowTransformAtPlacement: Singular matrix [nan 0.000 0.000 nan]
Jun 23 09:06:40 eta com.apple.Dock.agent[99877]: Thu Jun 23 09:06:40 eta.colorado.edu Dock[99877] : kCGErrorIllegalArgument: CGSSetWindowTransformAtPlacement: Singular matrix [nan 0.000 0.000 nan]

are correlated with opening Chrome windows and/or Chrome's crash_inspector

Jun 23 09:06:09 eta [0x0-0x69e69e].com.google.Chrome[99995]: [99995:24579:485131152128125:ERROR:shared_memory_posix.cc(164)] Creating shared memory in /var/folders/ni/ni+DtdqFGMeSMH13AvkNkU+++TI/-Tmp-/.com.google.chrome.sHcu6r failed: Too many open files in system

This is the problem that really gets me... I think it's crash_inspector's fault.

But there's definitely more going on here than just Chrome. Trying to change default browsers (by opening Safari and opening Preferences) led to a partial Dock crash (?!) in which I can alt-tab but can't see the Dock. Not clear at all what's going on.... argh.

Dying Dock

My dock keeps dying. Repeatedly. Over and over.

Only solution so far:

ps -vax | grep -E "Dock|PID"
kill -HUP PID
kill -s SIGCHLD PID


And similarly for problems with Chrome + /usr/sbin/mDNSResponder. They tend to go bad together.... no clues yet from the system logs. Ironically, the crash reporter seems to fail the most often...

specanpy

AG

There are now 2 clones of PySpecKit:
sPecAnPy
Spectroscopic Toolkit (Astronomy)

Why can't numpy do duplicate index assignment

AG

I want to do drizzling with numpy. It should be trivial, but it's impossible (without a for loop, afaik) instead.


In [2]: a = array([1,1,2,2])

In [3]: b = arange(5)

In [4]: b[a] += 1

In [5]: b
Out[5]: array([0, 2, 3, 3, 4])

In [6]: # but b should really be:

In [7]: b[a] += 1

In [8]: b
Out[8]: array([0, 3, 4, 3, 4])


pyspeckit: an astronomical spectroscopic toolkit

Jordan and I have been working on our python-based spectroscopic analysis tool for a while now:
pyspeckit is a pretty awesome, now functional but incomplete (and incompletely documented) tool.

pstopng

Following this thread and my need to convert IDL .ps files to .pngs so that I can view them in Mac Preview without having to go through an (often failed) conversion process led to a few discoveries.

First, it is challenging to get ImageMagick's convert to make an opaque background for the .ps file without seriously degrading the resolution. This can be accomplished by passing both the -alpha Off and the -density 300 simultaneously. This is slow, though, and recommendations to speed it up using the -limit area 4096 -limit memory 4096 tags actually made it slower!

However, the thread pointed out that ghostscript can do the conversion directly:
gs -dBATCH -sDEVICE=png16m -r300 -dEPSCrop -dNOPAUSE -sOutputFile=XXX.png XXX.ps

-sDEVICE sets the output to png, -r300 tag sets the density to be 300 pixels/inch, -dEPSCrop is necessary to get the right sized image out (otherwise it defaults to a portrait 8.5x11), and -dBATCH prevents the gs command line from activating after the command is executed. I'm not sure if -dNOPAUSE is necessary, but apparently if you don't activate it you have to do something after every page is processed.

My code to do batch ps-to-png conversion is available at http://code.google.com/p/agpy/source/browse/trunk/agpy/pstopng.

Timing demonstrations (units are seconds, R is 'real' or clock time, 'U' is user time, and 'S' is system time):

/usr/local/bin/convert -density 300 -alpha Off deline_zero_10hz_timestreams_003.ps deline_zero_10hz_timestreams_003.png
TIMING: R: 3.126 U: 2.948 S: 0.084
/usr/local/bin/convert -limit area 4096 -limit memory 4096 -density 300 -alpha Off deline_zero_10hz_timestreams_003.ps deline_zero_10hz_timestreams_003.png
TIMING: R: 3.800 U: 2.970 S: 0.161
gs -dBATCH -sDEVICE=png16m -r300 -dEPSCrop -dNOPAUSE -sOutputFile=deline_zero_10hz_timestreams_003.png deline_zero_10hz_timestreams_003.ps
TIMING: R: 0.801 U: 0.781 S: 0.017