RE: EPICS base 3.15.0.2

["Heesterman, Peter J" <Peter.Heesterman@ccfe.ac.uk>]

Hi Andrew,

My install process has been to delete my pre-existing 'base' directory, copy your download in its place, and then recover my non-downloaded files (i.e., principally my VS project files) over the top, from my source control server.

I have carefully compared - windiff - my install directory with your raw download directory.

Changed files are:

Due only to the patch you sent me earlier:
.\src\libcom\flex\makefile
.\src\libcom\flex\rules

For the reasons previously discussed:
.\src\libcom\misc\epicstypes.h
.\src\std\softioc\base.dbd

I have customised Win32.bat - I believe correctly - to suit my build environment and paths,  and have retained my version.
It has not changed from my builds prior to this download.

>> Make -v
I note that it says
	This program built for i386-pc-mingw32

I'm using GnuWin32, not mingw, for my build.
Does this make sense?

Cheers,

Peter.


-----Original Message-----
From: Andrew Johnson [mailto:anj@aps.anl.gov] 
Sent: 31 October 2014 17:16
To: Heesterman, Peter
Cc: rivers@cars.uchicago.edu; EPICS core-talk
Subject: Re: EPICS base 3.15.0.2

Hi Peter,

On 10/31/2014 06:23 AM, Heesterman, Peter J wrote:
> I have applied the patch, but this hasn't resolved the problem.
> 
> NB, flex.skel.static was only the first of many files that couldn't be 
> found. I had already (manually) copied the file - and a few more - but 
> gave up on this approach.

I suspect you may have something wrong with your build configuration or your tools setup then. Other than running 'make' our Makefile builds of EPICS Base should not require *any* manual operations. Are you building from a pristine copy of the 3.15.0.2 tarfile, or have you made some local changes, and what is your output from running 'make -v'?

You can see full build logs for the 3.15 branch at our Jenkins server https://jenkins.aps.anl.gov/view/EPICS%20Base/ so the win32-x86 log is at https://jenkins.aps.anl.gov/view/EPICS%20Base/job/epics-base-3.15-win32/lastBuild/consoleFull

I compared the logfile that you sent earlier in this thread line-by- line with one I generated here. The first major difference is that your build did not install any of the libCom include files, which includes the flex.skel.static file. Between lines 362 and 365 of your logfile I have an additional ~120 lines of installation messages.

>>> epicsBoolean
> I wondered if it would be reasonable to typedef epicsBoolean to bool 
> (rather than #define to int)?

Unfortunately we don't have C99 support and hence the bool typedef on all our target architectures, so that would break builds on those architectures. The Asyn uses of epicsBoolean are being replaced but I will leave the definition in now that I've restored it and will mark it as deprecated.

> BTW, I also noticed that if you type cd<RET> to the IOC shell, the 
> application falls over. This is due to null pointer dereference as no 
> directory string has been provided. (In the Windows shell, of course, 
> this usage prints out the current directory.)

Acknowledged; the Posix standard doesn't say if a NULL argument to
chdir() is legal, so we should avoid passing it one. I fixed this and a similar issue in the echo command, thanks.

- Andrew

--
People everywhere confuse what they read in newspapers with news.
-- A. J. Liebling

Attachment: makemv.log
Description: makemv.log