Experimental Physics and
Industrial Control System

"Jeff Hill" <[email protected]> · Wed, 23 Nov 2005 15:25:04 -0700

> But if I tried to build shared libraries, I
> must copy the manifest specification file for
> each exe or dll file into bin/win32-x86 manually.

I created mantis entry 231.

> I also found that I can not execute the catools such as
> caget/cainfo/camonitor correctly.

This appears to be related to this line in caget.

    setvbuf(stdout,NULL,_IOLBF,0);    /* Set stdout to line buffering */

Apparently the new ms CRTL assert fails 2 <= size <= INT_MAX. That's a new
feature of visual C++ 2005 (see
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vclib/html/
_crt_setvbuf.asp). This does not match POSIX (see
http://www.opengroup.org/onlinepubs/007908799/xsh/setvbuf.html).

I did find that making this change appears to fix the problem.

    setvbuf(stdout,NULL,_IOLBF,2);    /* Set stdout to line buffering */

I created mantis entry 230.

Jeff

> -----Original Message-----
> From: guobao shen [mailto:[email protected]]
> Sent: Wednesday, November 23, 2005 5:09 AM
> To: 'Jeff Hill'; 'Vladimir Sirotenko'; [email protected]
> Subject: RE: EPICS base compiling using Visual C++ 2005 Express
> 
> Dear Dr. Jeff and Dr. Vladimir,
> 
> Thank you very much.
> 
> I succeeded in the compilation after removing the TD from the fdopen
> specification.
> If I tried to build static libraries, there was no problem. But if I tried
> to build shared
> libraried, I must copy the manifest specification file for each exe or dll
> file into
> bin/win32-x86 manually.
> 
> I also found that I can not execute the catools such as
> caget/cainfo/camonitor correctly.
> The only error message is "caget.exe has encountered a problem and needs
> to
> close."
> 
> Can you give any suggestion?
> 
> Thanks again.
> 
> Best Regards,
> 
> Guobao Shen
> J-PARC Control Group
> 
> 
> >
> > I looked into this "w+bTD" verses "W+B".
> >
> > The purpose of this code is to open an identical type of
> > temporary file on all OS. This particular problem is, of
> > course, in Microsoft specific code working with previous
> > versions, but not with Visual C++ 2005.
> >
> > The very latest documentation at MS with a 2005 time stamp on
> > it lists "T"
> > and "D" as valid Microsoft specific options to _fdopen. This
> > therefore smells like a bug at MS. Nevertheless, the proposed
> > change appears to work - not creating any permanent files on
> > the disk. This is probably because roughly equivalent options
> > are specified to the preceding open().
> >
> > We have committed the suggested change - ETA R3.14.8.
> >
> > Jeff
> >
> > > -----Original Message-----
> > > From: Vladimir Sirotenko [mailto:[email protected]]
> > > Sent: Tuesday, November 22, 2005 9:34 AM
> > > To: guobao shen; [email protected]
> > > Subject: Re: EPICS base compiling using Visual C++ 2005 Express
> > >
> > > Try to do the following, it worked for us:
> > >
> > > Antelope failed to run correctly because it could not create a
> > > temporary file.
> > > Modify src\libCom\osi\WIN32\epicsTempFile.cpp and remove
> > the TD from
> > > the fdopen specification.
> > >
> > >         pNewFile = _fdopen ( fd, "w+b" ); instead of pNewFile =
> > > _fdopen ( fd, "w+bTD" );
> > >
> > > Vladimir
> > >

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System