3.7win without IPC fix #1852

bagong · 2016-02-04T04:41:44Z

Consider this a discussion base. It basically contains most from the old Winport branch except Lucas Cornelisses IPC fix. So it's a Win version that builds but has the bug.

I added:

Miguel's open long files PR.
tims commit containing most of Lucas' stuff for VS64bit compatibility (I "cherry-pick"ed that single commit from master. Not sure what else that brought in that wasn't planned to be in the 3.7 branch)
other smaller stuff from Lucas' port that allow the VS build to succeed (like an alternative to sys/time.h and some small Qt-related stuff
cmake changes/additions that make building with VS and msys simple
a README that is in sync with those cmake changes

99% of this is purely Windows related and doesn't effect the other OSes.

In theory somebody could try to take Lucas' IPC fix and apply it on a branch that branches from this release branch. That would allow for a Windows version that is still fairly close to 3.7, and the other OSes wouldn't be affected by the IPC fix... Unfortunately I cannot do this myself, but merging this would provide a useful base...

changed External control -> External Control so that it ends up in the right scdoc category

scdoc: typo in Working_with_HID.schelp

scdoc: typo in LID_permissions.schelp

scdoc: typo in HID_permissions.schelp

…ectures based on a patch by Lucas Cornelisse Signed-off-by: Tim Blechmann <tim@klingt.org>

crucialfelix · 2016-02-04T10:44:40Z

Awesome !

timblechmann · 2016-02-08T04:58:13Z

CMakeLists.txt

@@ -24,6 +24,19 @@ if (${CMAKE_SOURCE_DIR} STREQUAL ${CMAKE_BINARY_DIR})
 	message(WARNING "WARNING: IN-PLACE BUILDS ARE NOT RECOMMENDED - PLEASE USE A BUILD DIRECTORY")
 endif()

+# allows cmake-find modules and Windows installer to identify the right architecture
+if(WIN32)
+    if(CMAKE_GENERATOR MATCHES "64" OR "$ENV{MSYSTEM}" MATCHES "64")


CMAKE_CL_64 to detect 64bit msvc.

bagong · 2016-02-14T10:14:21Z

The Travis error is the same one as before...

As to the section on building SC this is in sync with the changes provided by this branch ;)

Hoping Tavis will like this

bagong · 2016-02-18T02:29:39Z

For some reason Travis hasn't started building that last commit. But according to Winmerge this branch is now identical with the testbranch I was building up... And there it builds.. Don't know how to trigger the build manually

bagong · 2016-02-18T16:44:46Z

I am done with this. It looks as Travis didn't build the last commit, but as far as I can see, it did, only it seems to be out of sync with my commits. The ad8d9dc (second last) commit is marked okay now, although that was marked failed when it was initially triggered. I did try to retrigger the pull-request manually and it stayed on the second last commit but succeeded. What I think happened is that it compiled the last commit but marked the second last commit. The last commit, according to winmerge is exactly identical with the last commit on the testTravis fake PR I made to see when the error actually comes up. It seems, the error is triggered when in cmake path expressions that contain variables are not wrapped in string delimiters. Not in every case, but in the ones I corrected. That python test is completely intransparent...
So I think I met all request by Tim

except the "feed my grandma cast problem", which I put in OS check if's to use it only when we compile with VS: bagong@39a766a
The Q_OS_XXX thing (bagong@0418264) required to reformulate the ifdefs as suggested by Tim. Apparently that's a Qt thing, I saw it somewhere on stackoverflow. I also include QtGlobal. Including QtGlobal only (without the syntactical adjustment) actually made the build break for MinGW too at this place, so it may be that this construction was simply ignored on some platforms.
The biggest thing is the portaudio integration. Note that this is completely ignored on other platforms than Windows. I tested on both Linux and Apple that their build system is not contaminated when building with portaudio. My consideration for the inclusion of PA: on Windows you cannot get binaries of PA. Furthermore it is really difficult to compile PA using cmake if you want to get anything more than the basic WMME support. The only way to build PA fairly easily on Windows with other sound-api's is using msys and the unix build tools. That again is a different computer culture which I now tried to make unnecessary for building SC on Windows. What happens with this setup is: if a user does not provide PS libraries, SC will build a basic version as part of the SC build. If he/she does provide a library, that library will be preferred (can static or dynamic). Something I didn't manage is to allow the in place/embedded build of pa compile anything else than WMME support, but: the pa source is patched in a way that if he or she build PA separately from th SC build, the cmake pa build is going to succeed for all audio API's except WASAPI (which is excluded for MinGW in the options). I puzzled together patches from 3 different sources until it succeeded. You can see the patches in the submodule, it's a git repo and the patches are commits. I am happy to move that git repo to some neutral place. It git svn's the portaudio svn repo and is at latest pa-svn
remains the master commits that came in with the big Tim commit which I cherry-picked from master. In my opinion we shouldn't be too formalistic: these are fixes of typos in documentation, right. What bad can come from them in the 3.7 branch? So the only relevant one is Dan's bagong@a630e05. Should I revert it, or is it actually good to have it in 3.7?

best
.r.

bagong · 2016-02-19T10:33:57Z

Lucas' IPC fix is also done. It's a single commit on top of this, nicely separated. So with that commit Windows is stable, remains MIDI and the interactive console from the obvious things...

bagong · 2016-02-21T14:29:08Z

Great! ;)

crucialfelix · 2016-02-21T14:37:03Z

ok, now done ? (you added a few commits yesterday)

As you say, the big thing is adding portaudio in. The submodule will be there even for non-windows builds. This is a downside. Its your own fork, and that may causes stress/maintenance issues later. But at least we have windows running now, and the alternatives are worse. It can be easily switched for another (possibly an pa official) version in the future.

That read me file is pretty epic. I didn't read it all. In the end do Anna Krenina and Count Vronsky live happily ever after ?

We should merge this now, oder ? If you want to wait a day or two just to be sure, that's fine too.

You are a hero @bagong !

bagong · 2016-02-21T14:40:20Z

Thanks! Yep, merge ;)
As to portaudio: as I stated, I am happy to move the PA submodule to the supercollider repo, I would actually prefer that...

3.7win without IPC fix

crucialfelix · 2016-02-21T18:15:47Z

If there is any need to change it or burden on you then we can move it into supercollider when that time comes.

bagong · 2016-02-21T18:40:37Z

okle

bagong · 2016-02-21T22:47:04Z

I have pushed Lucas' "IPCfix" that makes the Windows version stable as branch "topic/IPCwithQTcp. Depending on what happens until the release, it could be used to make the win release stable... I'll try to keep it on top of 3.7...

redFrik and others added 16 commits December 6, 2015 16:05

scdoc: typo in HID_permissions.schelp

8e30bd6

changed External control -> External Control so that it ends up in the right scdoc category

scdoc: typo in LID_permissions.schelp

bd579ca

changed External control -> External Control so that it ends up in the right scdoc category

scdoc: typo in Working_with_HID.schelp

4b5fc03

changed External control -> External Control so that it ends up in the right scdoc category

Merge pull request #1756 from redFrik/patch-11

7c086c6

scdoc: typo in Working_with_HID.schelp

Merge pull request #1755 from redFrik/patch-10

f1adfcd

scdoc: typo in LID_permissions.schelp

build: fix PanUGens to build when NOVA_SIMD=OFF

a630e05

Merge pull request #1754 from redFrik/patch-9

130095d

scdoc: typo in HID_permissions.schelp

scdoc/sclang lexers: use intptr_t for compatibility with LLP64 archit…

c0a550d

…ectures based on a patch by Lucas Cornelisse Signed-off-by: Tim Blechmann <tim@klingt.org>

ide: fix issue with large documents

f5b7843

Merge commit 'c0a550d247ef98088dba3ca12c0e49587c2b6748' into 3.7win

3f2da04

cmake additions for Windows build

77bb46f

Additional Windows-specific elements in Lucas C. winport

b8febf5

Mainly cmake related changes to allow for VS Windows build

bd89b5c

Two typos and two workarounds for compile VS related compile problems

0ef2726

Another small correction

2a98c12

Readme with build instructions. Quite wordy, sorrz....

04b1660

crucialfelix added this to the 3.7.0 milestone Feb 4, 2016

crucialfelix added the os: Windows label Feb 4, 2016

Use less accurate system time for Win XP and Win7 compatibility

69d42d6

timblechmann reviewed Feb 8, 2016
View reviewed changes

bagong added 2 commits February 14, 2016 05:44

Add submodule portaudio

468daf9

Integrate embedded portaudio into Windows build system

7cf7be4

miguel-negrao and others added 13 commits February 14, 2016 11:58

ide: fix memory leak

b62004b

Make dll-search path a bit nicer

f620dae

Rewritten README_WINDOWS

d27796a

As to the section on building SC this is in sync with the changes provided by this branch ;)

Msys2 finds portaudio2 - behaves like Apple

636d12b

Correct omission in last commit

acfda0d

Deploy Qt platform plugin with with generator expression

0bf8729

Use Qt syntax for ifdef/if defined and include QtGlobal

0418264

Restrict eat-my grandma workaround to Visual Studio compiles

39a766a

Another little omission

a78c6f0

Use string delimiters consistently when using variables in paths

92cc290

Correct wrong elseif

17ee397

Hoping Tavis will like this

More white-space and delimiters

ad8d9dc

Unify with comparisonBranch

9ae01dd

bagong added 2 commits February 20, 2016 18:01

Switch portaudio linking to shared by default

fb4e463

Add change to portaudio repo

552e8bd

telephon added a commit that referenced this pull request Feb 21, 2016

Merge pull request #1852 from bagong/3.7win

3ed7345

3.7win without IPC fix

telephon merged commit 3ed7345 into supercollider:3.7 Feb 21, 2016

bagong deleted the 3.7win branch February 21, 2016 22:03

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

3.7win without IPC fix #1852

3.7win without IPC fix #1852

bagong commented Feb 4, 2016

crucialfelix commented Feb 4, 2016

timblechmann Feb 8, 2016

bagong commented Feb 14, 2016

bagong commented Feb 18, 2016

bagong commented Feb 18, 2016

bagong commented Feb 19, 2016

bagong commented Feb 21, 2016

crucialfelix commented Feb 21, 2016

bagong commented Feb 21, 2016

crucialfelix commented Feb 21, 2016

bagong commented Feb 21, 2016

bagong commented Feb 21, 2016

3.7win without IPC fix #1852

3.7win without IPC fix #1852

Conversation

bagong commented Feb 4, 2016

crucialfelix commented Feb 4, 2016

timblechmann Feb 8, 2016

Choose a reason for hiding this comment

bagong commented Feb 14, 2016

bagong commented Feb 18, 2016

bagong commented Feb 18, 2016

bagong commented Feb 19, 2016

bagong commented Feb 21, 2016

crucialfelix commented Feb 21, 2016

bagong commented Feb 21, 2016

crucialfelix commented Feb 21, 2016

bagong commented Feb 21, 2016

bagong commented Feb 21, 2016