Experimental Physics and
Industrial Control System

"Mark S. Engbretson" <[email protected]> · Tue, 29 Sep 2015 12:37:32 -0500

Not wanting to sound like a total ass . . . . but these sorts of issues seem to pop up all the time in Tech-Talk when people want to deploy this software for the first time, or after versions changes on certain modules.  While validating the builds against those 4 different archs would be nice, could it also be extended to rebuilding all of this on totally fresh (virtual) systems? The #1 phrase that I total hate as an answer to anything is, "It works on my system."

Probably why people park on old versions of software for very long periods.

-----Original Message-----
From: Mark Rivers [mailto:[email protected]] 
Sent: Tuesday, September 29, 2015 12:17 PM
To: Kevin Peterson <[email protected]>; J. Lewis Muir <[email protected]>; Mark S. Engbretson <[email protected]>; [email protected]
Subject: RE: Unable to build support modules on Windows, with base 3.15.2

Folks,

I have determined that the following changes are required to build synApps_5_8 on win32-x86.

- areaDetector has the following problems:

  - The top-level areaDetector is R2-0, while ADCore is R2-2.
    This is not allowed, the RELEASE.md for ADCore R2-2 clearly states that
    it requires R2-2 of the top-level areaDetector.  It was just luck that this
    worked on Linux and vxWorks.
    The problem on Windows is that the CONFIG_SITE.local file is missing some 
    critical lines.  Adding the following lines will allow it to work.
ifeq ($(OS_CLASS), WIN32)
    HDF5_STATIC_BUILD=$(STATIC_BUILD)
endif

  - The supplied top-level configure/RELEASE will not work under Windows on a Linux
    file system.  The normal areaDetector RELEASE_PATHS.local, 
    RELEASE_PATHS.local.win32-x86, RELEASE_LIBS.local, and RELEASE_PRODS.local
    must be used.
    

- seq needs to be updated to seq-2-2-snapshot-2015-05-27.  seq-2-2-1 and seq-2-2-2 do not build on win32-x86.  Hopefully there will be a new official release of seq soon that builds on win32-x86.

- alive needs to be updated to the Subversion trunk version.  This should be tagged as 1-1 soon.

- caputRecorder needs to be updated to 1-4-3.  1-4-2 does not build on dynamic Windows architectures.

- softGlue needs to be updated to 2-5.  2-4-3 does not build on Windows.

- love-3-2-5 has an unneeded file configure/RELEASE.win32-x86 that makes the build fail because it is not updated by "make release".  Delete that file or update to Subversion trunk version.

- motor-6-9 will not build because of a problem in PhytronSrc.  Edit
  motorApp/PhytronSrc/phytronMotorAxis.cpp to make the following patch:
+#ifndef _WIN32
 #include <unistd.h>
+#endif

- The allenBradley, devIocStats, ipac, and seq modules cannot be built under Windows on a Linux file system until the following line is added to configure/RELEASE immediately following the line that defines EPICS_BASE:
-include $(TOP)/../configure/EPICS_BASE.$(EPICS_HOST_ARCH)

devIocStats/configure/RELEASE also needs the following line right after the line that defines SUPPORT:
-include $(TOP)/../configure/SUPPORT.$(EPICS_HOST_ARCH)

xxx needs the following changes to configure/RELEASE:
corvette:support/xxx-5-8-3/configure>diff -U3 ~/scratch/synApps_5_8/support/xxx-5-8-3/configure/RELEASE .

--- /home/epics/scratch/synApps_5_8/support/xxx-5-8-3/configure/RELEASE	2015-03-27 13:45:36.000000000 -0500
+++ ./RELEASE	2015-09-29 11:12:19.600114581 -0500
@@ -26,10 +26,13 @@
 ##############################################################################
 
 # Define the path to the synApps support directory and uncomment.
SUPPORT=/home/oxygen/MOONEY/distrib/synApps_5_8/support
+-include $(TOP)/../configure/SUPPORT.$(EPICS_HOST_ARCH)
 
 # Specify the standard synApps configuration.
 include $(SUPPORT)/configure/RELEASE
+-include $(TOP)/../configure/SUPPORT.$(EPICS_HOST_ARCH)
+-include $(TOP)/../configure/EPICS_BASE.$(EPICS_HOST_ARCH)
 #CAPUTLOG=$(SUPPORT)/caPutLog-3-4


I think the synApps_5_8 Known Problems page should contain the above information.

I also think that in the future synApps should be tested on the following 4 architectures before a new release is issued: win32-x86, win32-x86-static, windows-x64, windows-x64-static.

Mark


________________________________________
From: Kevin Peterson [[email protected]]
Sent: Thursday, September 24, 2015 1:54 PM
To: Mark Rivers; J. Lewis Muir; Mark S. Engbretson; [email protected]
Subject: Re: Unable to build support modules on Windows, with base 3.15.2

I believe I get the following link error when building seq-2-2-1 (included with synApps_5_8) and seq-2-2-2 when building for win32-x86:

link -nologo  /LTCG /incremental:no /opt:ref /release
-out:snc.exe        lexer.obj snl.obj main.obj node.obj var_types.obj
analysis.obj gen_code
.obj gen_ss_code.obj gen_tables.obj sym_table.obj builtin.obj
type_check.obj     ..\\..\\..\\lib\\win32-x86\\seq.lib
..\\..\\..\\lib\\win32-x86\\pv.lib  D:\\ep ics\\base-3.14.12.5\\lib\\win32-x86\\ca.lib
D:\\epics\\base-3.14.12.5\\lib\\win32-x86\\Com.lib
var_types.obj : error LNK2001: unresolved external symbol __imp__prim_type_name gen_tables.obj : error LNK2001: unresolved external symbol __imp__prim_type_tag_name snc.exe : fatal error LNK1120: 2 unresolved externals
make[4]: *** [snc.exe] Error 1120
make[4]: Leaving directory
`D:/epics/synApps_5_8/support/seq-2-2-2/src/snc/O.win32-x86'
make[3]: *** [install.win32-x86] Error 2
make[3]: Leaving directory `D:/epics/synApps_5_8/support/seq-2-2-2/src/snc'
make[2]: *** [snc.install] Error 2
make[2]: Leaving directory `D:/epics/synApps_5_8/support/seq-2-2-2/src'
make[1]: *** [src.install] Error 2
make[1]: Leaving directory `D:/epics/synApps_5_8/support/seq-2-2-2'
make: *** [D:/epics/synApps_5_8/support/seq-2-2-2.install] Error 2

alive-1-0 doesn't build for Windows unless you add the following libs to its src/Makefile.  The alive svn repo contains a similar fix, but it assumes ws2_64 is available, which it wasn't on my system.

+alive_SYS_LIBS_WIN32 += ws2_32
+alive_SYS_LIBS_WIN32 += Advapi32

The version of ADCore included in synApps_5_8 fails to link out-of-the-box for win32-x86:

link /nologo /subsystem:windows /dll /LTCG /incremental:no /opt:ref
/release        /implib:NeXus.lib /out:NeXus.dll          napi.obj
napi5.obj napiu.obj nxdat
aset.obj nxio.obj nxstack.obj nxxml.obj stptok.obj D:\\epics\\synApps_5_8\\support\\areaDetector-R2-0\\ADBinaries-R2-2\\lib\\win32-x86\\libhdf5.lib
 D:\\ep
ics\\synApps_5_8\\support\\areaDetector-R2-0\\ADBinaries-R2-2\\lib\\win32-x86\\libszip.lib
 D:\\epics\\synApps_5_8\\support\\areaDetector-R2-0\\ADBinaries-R2-2\
\lib\\win32-x86\\libzlib.lib
LIBCMT.lib(timeset.obj) : error LNK2005: ___daylight already defined in
MSVCRT.lib(MSVCR100.dll)
LIBCMT.lib(timeset.obj) : error LNK2005: ___timezone already defined in
MSVCRT.lib(MSVCR100.dll)
LINK : warning LNK4098: defaultlib 'LIBCMT' conflicts with use of other libs; use /NODEFAULTLIB:library NeXus.dll : fatal error LNK1169: one or more multiply defined symbols found
make[5]: *** [NeXus.dll] Error 1169
make[5]: Leaving directory
`D:/epics/synApps_5_8/support/areaDetector-R2-0/ADCore-R2-2/ADApp/nexusSrc/O.win32-x86'
make[4]: *** [install.win32-x86] Error 2
make[4]: Leaving directory
`D:/epics/synApps_5_8/support/areaDetector-R2-0/ADCore-R2-2/ADApp/nexusSrc'
make[3]: *** [nexusSrc.install] Error 2
make[3]: Leaving directory
`D:/epics/synApps_5_8/support/areaDetector-R2-0/ADCore-R2-2/ADApp'
make[2]: *** [ADApp.install] Error 2
make[2]: Leaving directory
`D:/epics/synApps_5_8/support/areaDetector-R2-0/ADCore-R2-2'
make[1]: ***
[D:/epics/synApps_5_8/support/areaDetector-R2-0/ADCore-R2-2.install] Error 2
make[1]: Leaving directory `D:/epics/synApps_5_8/support/areaDetector-R2-0'
make: *** [D:/epics/synApps_5_8/support/areaDetector-R2-0.install] Error 2

Those are the few errors that I copied into a text file before I gave up trying to build synApps_5_8 for its areaDetector support.

Kevin

On 9/24/15 1:35 PM, Mark Rivers wrote:
>> Many of the problems seem to be caused by Makefiles that 
>> unconditionally build code that doesn't work on Windows, and 
>> therefore are a problem for all versions of Visual Studio.
>
> I build all of synApps, including the sequencer on Windows, and don't have those problems.  What modules specifically did you have trouble with?
>
> Mark
>
>
> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Kevin Peterson
> Sent: Thursday, September 24, 2015 1:33 PM
> To: J. Lewis Muir; Mark S. Engbretson; [email protected]
> Subject: Re: Unable to build support modules on Windows, with base 
> 3.15.2
>
> On 9/24/15 10:45 AM, J. Lewis Muir wrote:
>>> If you succeed at building base, you'll encounter more problems when 
>>> you try to build other EPICS modules.
>>
>> Which problems are these?  If a certain compiler does not build EPICS 
>> software correctly, it would be good to note that compiler as 
>> unsupported on Andrew's page.
>
> I have not had a chance to document all of the problems I encountered 
> trying to build synApps_5_8 with Visual Studio Community 2015.  Many 
> of the problems seem to be caused by Makefiles that unconditionally 
> build code that doesn't work on Windows, and therefore are a problem 
> for all versions of Visual Studio.
>
> The sequencer is one module that is difficult to build on Windows, but 
> those problems have already been discussed here.
>
> Kevin
>





Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System