Until someone figures out what the hack really means [ dmg.grow(...) ] I
wouldn't consider it a patch. Clues can be found in
rep/binject/ext/binject_c/binject.c and hfslib.c (the grow_hfs
function). Without knowing much about that code, I can only guess that
its the max size of the .dmg - it could also be the minimum for all I
know and using a high value might be wasting space and larger downloads
for simple apps. Since we're in C code, that explains the crash instead
of a ruby trace. 

Someone who really knows the code will have to make those changes. I'm
not comfy in there. 

Since a lot of pack.rb is a Shoes page/applet one could add a gui field
to specify the size with a default to the 10MB limit for the rare
occasion someone packs in a lot of stuff. Better yet, a clever person
could write code to better guess the proper value.  Won't be me. 


--Cecil

On Sat, 2009-05-23 at 16:28 -0700, Noah Thorp wrote:
> I wonder if there are side effects to setting this value rather than  
> sensing the size of the application. I have a github account... just  
> not sure if it is best as a trunk or local fix without knowing much  
> about the internals of shoes.
> 
> -Noah
> 
> On May 23, 2009, at 3:48 PM, e deleflie wrote:
> 
> >> Hey Cecil,
> >>
> >> Thank you! The hack you suggested worked I now have a dmg file which
> >> includes the 10mb ffmpeg file.
> >
> > great!
> >
> > Assuming that _why doesn't read all messages in all threads, how can
> > we make sure this fix gets put into the source ... is it all about
> > github? Does someone have an account on github that they can push this
> > through?
> >
> > Etienne
> >
> >> Yes, ffmpeg wrappers are a major pain with all the targets. The  
> >> biggest pain
> >> though is the licensing problem for encoders - I am still sorting  
> >> out the
> >> legal implications of distributing aac and h264 encoders.  So far  
> >> it looks
> >> murky and not encouraging. ffmpegX gets around this by having you  
> >> download
> >> ffmpeg from some other site and can't legally automate this  
> >> process. I think
> >> that removes the simplicity that I want to deliver to the user.  
> >> Blech, I
> >> hate licensing quagmires!
> >>
> >> I just built a quick shoes prototype of a 1 click transcoder.  If  
> >> the the
> >> legal stuff doesn't get in the way, the idea is to have a batch  
> >> encoder with
> >> a configurable yaml file with ffmpeg parameters. It will upload  
> >> files to a
> >> website once the transcoding is done. The idea being that it does the
> >> conversion locally to one or several formats and can post the files  
> >> to a
> >> designated web service. This is instead of requiring the user to  
> >> upload
> >> gigabytes of data and it eliminates (or reduces) the need for a  
> >> transcoding
> >> farm. I think it would be benificial for lots of small sites that  
> >> require
> >> transcoding for their initial content, although it's usefulness  
> >> peters out
> >> after you change your mind about what formats you need.
> >>
> >> Anyway... I'll stop there since it's not really that super shoes  
> >> related.
> >>
> >> If you want to chat more about the ffmpeg stuff email me.
> >>
> >> Best,
> >> Noah
> >>
> >> On May 22, 2009, at 10:39 PM, Cecil Coupe wrote:
> >>
> >>> It's difficult meta problem. I'm on Linux (Ubuntu 9.04) and I'd  
> >>> like a
> >>> little shoes app that front ends ffmpeg. Except.... it has to  
> >>> front end
> >>> the ffmpeg I have, not the one someone thinks I need. Cross platform
> >>> reaching out to system installed dynamic libraries from inside  
> >>> Shoes is
> >>> a problem I shudder to think about. Mostly, Shoes won't. Hint's  
> >>> are in
> >>> the installer code. For good reasons it maybe shouldn't. Bundling  
> >>> your
> >>> own ffmpeg is also a big problem if you cover all the architectures,
> >>> that's no fun either.
> >>>
> >>> ffmepgX was (is) an awesome GUI wrapper. I still miss it on Linux. A
> >>> Shoes version would be a fine achievement. Keep me (or the list)
> >>> informed.
> >>>
> >>> --Cecil
> >>>
> >>> On Fri, 2009-05-22 at 21:37 -0700, Noah Thorp wrote:
> >>>>
> >>>> Cecil. Thanks for the lead I will have a look at that! I agree with
> >>>> you that I would prefer to find the bug rather than downloading  
> >>>> files
> >>>> - seem cleaner. I will have a look and see how it goes. It might  
> >>>> lead
> >>>> me to automating other aspects of the build process as well once  
> >>>> I see
> >>>> how the packaging works.
> >>>>
> >>>> Thanks!
> >>>> Noah
> >>>>
> >>>> On May 22, 2009, at 8:44 PM, Cecil Coupe wrote:
> >>>>
> >>>>> I hate to follow up my own message but
> >>>>>
> >>>>> pack.rb line 156 (my older version of shoes) has this line of ruby
> >>>>> dmg.grow(10)
> >>>>>
> >>>>> That seems like an arbitrary number that some folks have  
> >>>>> reported. I
> >>>>> don't know what dmg.grow() does, but a hacker might up that to  
> >>>>> 30 or
> >>>>> 40
> >>>>> and see what happens. Or one could read the docs for binject.  
> >>>>> Where
> >>>>> that
> >>>>> pack.rb is on your system and which copy of shoes you are using  
> >>>>> if you
> >>>>> have multiples...
> >>>>>
> >>>>> Just something to look at.
> >>>>> --Cecil
> >>>>>
> >>>>>
> >>>>> On Fri, 2009-05-22 at 21:00 -0600, Cecil Coupe wrote:
> >>>>>>
> >>>>>> It might be easier to debug and fix the problem than implement a
> >>>>>> workaround, IMHO. It looks like most of the action starts in
> >>>>>> lib/shoes/pack.rb - I don't have an OS X box anymore so I can't  
> >>>>>> be
> >>>>>> very
> >>>>>> helpful. Might be an easy fix if you can find all the parts.
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> On Fri, 2009-05-22 at 21:24 -0500, Roy Wright wrote:
> >>>>>>>
> >>>>>>> Maybe as a work around, implement a download and install  
> >>>>>>> ffmpeg as
> >>>>>>> needed instead of including the binary in the package?   
> >>>>>>> Similar to
> >>>>>>> how a
> >>>>>>> shoes app can download and install shoes...
> >>>>>>>
> >>>>>>> HTH,
> >>>>>>> Roy
> >>>>>>>
> >>>>>>> Noah Thorp wrote:
> >>>>>>>>
> >>>>>>>> Thanks for the encouragement. I have plans to post the open  
> >>>>>>>> source
> >>>>>>>> transcoder (ffmpeg wrapper) once it had reaches a certain level
> >>>>>>>> of alpha
> >>>>>>>> maturity. It's a very simple but useful app. I could post it to
> >>>>>>>> github
> >>>>>>>> early if it would help troubleshooting. But, it seems like it  
> >>>>>>>> may
> >>>>>>>> be an
> >>>>>>>> issue simply with including large files in an OS X package.
> >>>>>>>>
> >>>>>>>> Thanks,
> >>>>>>>> Noah
> >>>>>>>>
> >>>>>>>> On May 22, 2009, at 5:31 PM, e deleflie wrote:
> >>>>>>>>
> >>>>>>>>> Hi Noah,
> >>>>>>>>>
> >>>>>>>>> I dont think you are falling outside of the use case it was
> >>>>>>>>> intended
> >>>>>>>>> for ... I think its just a bug.
> >>>>>>>>>
> >>>>>>>>> I dont have a work-around for you .... (maybe someone else
> >>>>>>>>> does) ...
> >>>>>>>>> but if we shout loud enough .... _why might hear our plea to
> >>>>>>>>> have a
> >>>>>>>>> look at this bug.
> >>>>>>>>>
> >>>>>>>>> So it looks like large binaries (5 to 10MB and up?) included  
> >>>>>>>>> in a
> >>>>>>>>> folder and packaged up, for OSX, cause the packager to crash.
> >>>>>>>>>
> >>>>>>>>> Etienne
> >>>>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>
> >>>>
> >>>
> >>
> >>
> 

Reply via email to