win32 patches for backend, HDHR, datadirect, transcode

[Jeff Lu <jll544@…>]

The attached patches enable core backend functionality for win32. With these applied, a windows backend can scan channels and record from an HDHomeRun, download listings from SchedulesDirect?, and transcode recordings. There are still bugs, no doubt, but I have run the port as both standalone and slave backends with good results. There are four patches:

backend-win32.patch - allows mythtv-setup to detect backend running, fixes creation of temp directories, enables mythsystem calls to get exit codes, and fixes recording file corruption.

hdhr-win32.patch - minor updates to allow HDHR support to compile and run on win32.

datadirect-win32.patch - various updates to handle windows command line syntax.

transcode-win32.patch - enable mythtranscode to compile on win32, use QFile for file renaming (MinGW's own routines don't seem to handle filename extensions > 3 chars), and extend the existing hacks to prevent files being open after the transcode finishes.

[davidbuzz@…]

HI Jeff, visual review of these looks good!.

some possible improvements might include:

backend patch relies on the %PATH% being set correctly. If a unix/mingw/msys/cygwin find.exe command is in the path before the default windows one, then this breaks the backend code supplied. may I suggest full path definition of %SystemRoot?$/System32/find.exe be used.

From memory, I thin a lot of win32 code in mythtv uses #ifdef _WIN32 instead of #ifdef USING_MINGW. No biggie either way, I'm just thinking of standardising on one of them

A lot of single Vs double quoting issues - unix = ' windows = \" Perhaps using a QString.replace finction to convert single quote to backquoted double quotes on-the-fly on windows only might be more future-proof.?

str.replace(QString("'"), QString("
"")); (or similar )

A bunch of functions in NuppelVideoRecorder?.cpp are basically a no-op under mingw. Rather than #ifndef USING_MINGW inside all these functions, maybe you could supply these MINGW versions as a separate implementation, or move the null-op-mingw-versions to the end of the file? not sure.hhmmmm.

like I said though, the rest of it looks good from here ( untested). Buzz.

[Jeff Lu <jll544@…>]

Replaces previous patch; uses explicit paths; adds mythsocket fix

[Jeff Lu <jll544@…>]

Replaces previous DD patch; simplified to use QStrring.replace

[Jeff Lu <jll544@…>]

Thanks to David for the good suggestions. I've attached a couple revised patches: Backend patch now uses full paths and adds a new fix for mythsocket not setting socket buffer size for incoming connections. Also, the datadirect patch is now much simpler by using string replace.

For now, I've left the #ifdef's alone, as I seem to be seeing USING_MINGW more often in the code than _WIN32. I agree it would be best to standardize on one, though.

[Jeff Lu <jll544@…>]

Oops - Ignore previous DD patches. This is the right one.

[Jeff Lu <jll544@…>]

Replaces previous transcode patch; adds fixes for corruption and unclosed filedescriptor for lossless MPEG2

[Jeff Lu <jll544@…>]

Updated to r20837

[Jeff Lu <jll544@…>]

Updated to r21056

[Jeff Lu <jll544@…>]

Updated to r21056

[markk]

(In [21543]) win32 fixes for mythbackend and hdhomerun support. Refs #6597. Not extensively tested but I have encountered no obvious MythTV issues setting up a backend on Vista using a dvb-t hdhomerun. N.B. For some reason Vista does not ask if you want to allow mythtv-setup through the firewall (as it does for the frontend and backend). I had to manually add it to get hdhomerun channel scanning working.

[stuartm]

The datadirect patch is ok, but I'd like to see if we can't use httpcomms instead and remove the need for these platform specific fixes. If no-one wants to do that for 0.22 then we can use the patch for now and I'll look at it after the release.

[markk]

(In [21545]) win32 fixes for the Schedules Direct code. Refs #6597. The wget dependancy will likely be removed for 0.23.

[markk]

Push the last, uncommitted patch for mythtranscode to 0.23.

I'm getting numerous compilation issues and there are more general, unresolved issues related to PlayerContext? that need addressing first. As David noted, the numerous #ifdef's in NuppelVideoRecorder? should probably be handled more elegantly as well.

[markk]

Push back (again) to 0.24

Jeff has stated that he is no longer maintaining the transcode patch but I'd rather not lose sight of it.

[stuartm]

Milestone 0.25 deleted

[markk]

I can't see myself actually doing anything with the last transcode patch and it is relatively invasive, so marking as fixed for the other patches applied.