[mythtv] Re: [mythtv-commits] mythtv commit: r7932 by nigel

[George Nassas]

On 20-Nov-05, at 7:27 PM, Nigel Pearson wrote:

> 	Thanks for the improvements to the script -
> automating it does have some merit. Some discussion:
>
> 1) This checkout problem will probably be fixed soon.
>    Not sure of the merit of adding this much code in.

Sounds like you're given a good reason in 2) below :)

Anyway it's not much code. Three short functions and two brief ifs.

> 2) If we do add it, we should probably image everything.
>    Just imaging mythplugins will only fix the current problem,
>    and not some possible future case clashes in third party,
>    myththemes, or mythtv source.

I imaged just the plugins for performance reasons. An extra level of 
indirection can't be good, particularly for disk intensive stuff like 
linking. On the other hand most of the build is probably cpu-bound.

The other possibility is if someone knows how to detect filesystem 
properties on OSX I could make the script sense when an image is needed 
and take appropriate action. No switch needed.

> 3) In the unlikely event of 10.2 builds, or that an HFSX FS.
>    isn't good enough for some potential future problem,
>    we should probably use UFS?

My thought was an Apple-designed filesystem would be most efficient on 
an Apple OS but on the other hand HFSX does journalling which is no 
help on a throwaway disk image. I tried a build with UFS and it works 
so we can go with that.

> 4) "casehack" isn't the best name. How about "usehdimage"
>    or "buildinufs"

Sure, usehdimage sounds right to me. An autosense would be ever so much 
cooler.

 From the earlier posts:

> 	Until we get a "proper" solution,
> I have just put this in previewgenerator.cpp :
> #ifdef USING_BACKEND

Enabling the backend works and is probably a good idea if it's getting 
time to start packaging the backend for OSX. Is anyone working on this? 
I could give it a go in December sometime.

>> 3) when @svnrevision is empty perl's system function passes a null 
>> argument to svn which it doesn't appreciate at all.
>
> 	Hmmm. I never got this problem, and still don't.
> What error did you get back?
> Does the current version (er, r7943) exhibit the problem?

Odd, I don't get this today although it stopped me dead yesterday. The 
error was from subversion and it was "svn: ' ' does not appear to be a 
URL" and the verbose output showed an extra space between co and the 
myth svn url. If I moved the @svnrevision arg to the end of the svn 
checkout array then the checkout went OK and svn gave the error after 
that when it went to process the next args. Clearly an empty 
@svnrevision was causing trouble but now it's not. Hmph. I haven't 
upgraded anything. Oh well.

I'll make the usehdmimage changes and resubmit.

One other change I made affects how, in the absence of a revision or 
tag argument, the script does an svn update before building each 
component. Oftentimes there will be a checking during the mythtv build 
and the plugins will be from a different revision. The odds of a 
problem are small but why court trouble. For a HEAD build I've got it 
picking the revision at the beginning and using that for all the 
components.

Also, if you're just doing local hacking and you want to build a new 
frontend what are you supposed to do? The nohead seems to be a step in 
that direction but I don't understand all the cleaning of makefiles and 
reconfiguring. I mean if I just tweak one .cpp it should rebuild that, 
relink and repackage. How are we intended to do that?

- George