Hi Fons,
Thanks a lot for your reply.
Yep, I think it is a good idea to make a windows release in which
everything is linked against libNew, so that at least one has again
a fully functional ROOT version on windows.
I volunteer to test such a version on my win98 box with all my
analysis software. Once this works, one can indeed take the time
to go through the win32 interface and fix the uninitialised vars.
I understood Dave is coming to CERN soon for 3 months; maybe he
can give a hand in this effort.
Just let me know when I can get the 'libNew linked' ROOT binaries
from the web and I'll test them out.
Cheers,
Nick.
--
Dr. Nick van Eijndhoven mailto:nick@phys.uu.nl http://www.phys.uu.nl/~nick
--------------------------------------------------------------------------
Org.: Utrecht University/Nikhef, Department of Subatomic Physics
Address: P.O. Box 80.000, NL-3508 TA Utrecht, The Netherlands
Phone: +31-30-2532331(direct) +31-30-2531492(secr.) Fax: +31-30-2518689
NIKHEF: +31-20-5922028(direct) +31-20-5922000(secr.) Fax: +31-20-5925155
CERN: +41-22-7679751(direct) +41-22-7675857(secr.) Fax: +41-22-7679480
Offices: Buys Ballot lab. 710 (Utrecht) H350 (Nikhef) B23 1-020 (CERN)
--------------------------------------------------------------------------
Fons Rademakers wrote:
>
> Hi Nick,
>
> all problems you describe between the two version are due to the fact
> that the old version had all dll's linked with libNew. One of the things
> libNew is doing is to initialize all memory to 0 before objects are created.
> Recently I had to debug some other win32 issues and I found and corrected
> two non initialized variables in the win32 interface. I am sure there are
> more and they are the cuase of the problems. We just have to fix them one
> by one. On linux/unix we run often without libNew and there all known
> un-initialized variable problems have been fixed. Also the nice thing there
> is that you can activate libNew in any application and for all libs by just
> linking your app with or without libNew. On windows that is not possible.
> You have to link each DLL with libNew if you want to use it everywhere.
> This would force us to actually have to distributions if we want to give
> people the option to use either with or without them. I'll try to modify
> the makefile so that I can build with libNew for all DLL and we can see
> if the problems go. Later we still have to fix the uninitialized variables.
>
> -- Fons
>
> Nick van Eijndhoven wrote:
> >
> > Hi Anton,
> > There should be a simple solution, since ROOT 2.22/10 works correctly
> > on win98. Only later versions, in which the dependencies and lib creations
> > have been changed, show problems. So, somehow I get the feeling that in
> > making the change from 2.22 to 2.23 the problems have been introduced.
> > Furthermore, I remember that the first 2.23 release even crashed when one
> > just executed the command root.exe. This problem disappeared when I got
> > a test version of 2.23 from Valery Fine which he had created in the 'old cmz'
> > way. Afterwards the root team found and fixed the problem and released the
> > 2.23 version also for windows in the 'new fashion'. However, since that
> > time I have encountered several problems on win98.
> > This gives me the feeling that there might be a basic error in the ROOT
> > code somewhere (don't ask me where, maybe related with libNew) which
> > currently shows up on win98, but which might lateron also start to
> > cause problems on other operating systems. In view of this I would say
> > that it is better to look into the win98 problems and try to understand
> > and fix them, since also for other operating systems one might benefit
> > from it.
> > Cheers,
> > Nick.
> >
> > --
> > Dr. Nick van Eijndhoven mailto:nick@phys.uu.nl http://www.phys.uu.nl/~nick
> > --------------------------------------------------------------------------
> > Org.: Utrecht University/Nikhef, Department of Subatomic Physics
> > Address: P.O. Box 80.000, NL-3508 TA Utrecht, The Netherlands
> > Phone: +31-30-2532331(direct) +31-30-2531492(secr.) Fax: +31-30-2518689
> > NIKHEF: +31-20-5922028(direct) +31-20-5922000(secr.) Fax: +31-20-5925155
> > CERN: +41-22-7679751(direct) +41-22-7675857(secr.) Fax: +41-22-7679480
> > Offices: Buys Ballot lab. 710 (Utrecht) H350 (Nikhef) B23 1-020 (CERN)
> > --------------------------------------------------------------------------
> >
> > Anton Fokin wrote:
> > >
> > > Well,
> > >
> > > Once Fons said (with sarcasm though) that they perhaps should sell ROOT for
> > > windows as everybody does in the windows community and since windows itself
> > > is a commercial os. Perhaps this is right the solution. Instead of having
> > > lot of holes in the win distribution: missing GUI, threads, installations,
> > > crashes and so on and still trying to maintain win version, why don't you
> > > start taking money and employ _real_ windows experts. I mean you can not get
> > > _real_ windows experts for free. That's what we've been dealing with during
> > > last several years. _Real_ windows experts work for money: why on the Earth
> > > one should learn windows to write free soft for ROOT? Or should we give up
> > > windows distribution?
> > >
> > > Regards,
> > > Anton
> > >
> > > ----- Original Message -----
> > > From: Nick van Eijndhoven <Nick@phys.uu.nl>
> > > To: Anton Fokin <anton.fokin@smartquant.com>
> > > Cc: roottalk <roottalk@pcroot.cern.ch>; Fons Rademakers
> > > <Fons.Rademakers@cern.ch>; <dcasper@uci.edu>
> > > Sent: Tuesday, January 23, 2001 3:43 PM
> > > Subject: Re: [ROOT] Making test example on win98
> > >
> > > > Hi Anton,
> > > > It is indeed the crashes etc... w.r.t. some canvas actions which I am
> > > referring to.
> > > > For instance, if you have a histo and rightclick on one of the axes in
> > > order to
> > > > enter an axis title, the canvas disappears as soon as you have entered the
> > > text.
> > > > In the browser one can still see the canvas, but the only way of getting
> > > the
> > > > plot (in which indeed the axis text appears correctly) back again is to
> > > create
> > > > a new canvas and draw the histo again in that canvas.
> > > > Setting selection criteria by rightclicking in a TTreeViewer window
> > > however
> > > > directly results in a crash of the ROOT session.
> > > > Note that TTreeViewer also automatically opens 2 windows instead of only
> > > 1.
> > > > The latter has been observed also by the ROOT developers as far as I have
> > > understood
> > > > but so far the problem hasn't been fixed yet.
> > > >
> > > > In creating and executing my private programs and libs (i.e. 'batch'
> > > work),
> > > > everything works fine with the latest versions as far as I can see.
> > > > So, it is mainly the interactive analysis work which can not be performed
> > > > with version later than 2.22/10 on win98 machines, but needless to say
> > > > this is a very important feature which needs to function properly.
> > > >
> > > > Cheers,
> > > > Nick.
> > > >
> > > > --
> > > > Dr. Nick van Eijndhoven mailto:nick@phys.uu.nl http://www.phys.uu.nl/~nick
> > > > --------------------------------------------------------------------------
> > > > Org.: Utrecht University/Nikhef, Department of Subatomic Physics
> > > > Address: P.O. Box 80.000, NL-3508 TA Utrecht, The Netherlands
> > > > Phone: +31-30-2532331(direct) +31-30-2531492(secr.) Fax: +31-30-2518689
> > > > NIKHEF: +31-20-5922028(direct) +31-20-5922000(secr.) Fax: +31-20-5925155
> > > > CERN: +41-22-7679751(direct) +41-22-7675857(secr.) Fax: +41-22-7679480
> > > > Offices: Buys Ballot lab. 710 (Utrecht) H350 (Nikhef) B23 1-020 (CERN)
> > > > --------------------------------------------------------------------------
> > > >
> > > >
> > > > Anton Fokin wrote:
> > > > >
> > > > > Hi Nick,
> > > > >
> > > > > actually I am not creating ROOT binaries from scratch but rather trying
> > > to
> > > > > compile my own (unix) project on win98. What are the problems around
> > > root
> > > > > 3.02 bins on win98? I've been quite successful with making my libs under
> > > > > win98 and running exe and scripts, although I have noted that root
> > > crashes
> > > > > on some actions with canvas etc.
> > > > >
> > > > > PS. to "installshield" discussion. I checked out InstallShield coming
> > > with
> > > > > MSVC... It has "change registry" part in the install script, so what's
> > > the
> > > > > problem with setting paths under install? On the other hand what
> > > prevents
> > > > > you from writing a few lines in the shield script which will manually
> > > add
> > > > > two lines in autoexec.bat file? You have to restart com then but who
> > > > > cares...
> > > > >
> > > > > Regards,
> > > > > Anton
> > > > >
> > > > > ----- Original Message -----
> > > > > From: Nick van Eijndhoven <Nick@phys.uu.nl>
> > > > > To: Anton Fokin <anton.fokin@smartquant.com>
> > > > > Cc: roottalk <roottalk@pcroot.cern.ch>; Fons Rademakers
> > > > > <Fons.Rademakers@cern.ch>; <dcasper@uci.edu>
> > > > > Sent: Tuesday, January 23, 2001 12:55 PM
> > > > > Subject: Re: [ROOT] Making test example on win98
> > > > >
> > > > > > Hi Anton,
> > > > > > I am pleased to hear that you are creating ROOT binaries from scratch
> > > > > > on a Win98 platform.
> > > > > > As you know after ROOT 2.22/10 I have several problems when running
> > > the
> > > > > > ROOT binaries from the web on my win98 machine (you can ask Valery
> > > Fine
> > > > > > for details if you want). So win98 users are basically bound to stay
> > > with
> > > > > > the old 2.22/10 version.
> > > > > > I would like to suggest that once you have succeeded to create a
> > > > > completely
> > > > > > working set of ROOT binaries on a win98 system, that this version will
> > > > > also
> > > > > > be made available to the outside world via the ROOT web pages.
> > > > > > On the pages one could then indicate this version as win98 specific.
> > > > > > Could the ROOT team please consider this, since in our group many
> > > people
> > > > > > use laptops and for these things we have the feeling that the flexible
> > > > > win98
> > > > > > OS is much more convenient than winNT or win2000.
> > > > > >
> > > > > > Cheers,
> > > > > > Nick.
> > > > > >
> > > > > > --
> > > > > > Dr. Nick van Eijndhoven mailto:nick@phys.uu.nl
> > > http://www.phys.uu.nl/~nick
> > > > >
> > > > --------------------------------------------------------------------------
> > > > > > Org.: Utrecht University/Nikhef, Department of Subatomic Physics
> > > > > > Address: P.O. Box 80.000, NL-3508 TA Utrecht, The Netherlands
> > > > > > Phone: +31-30-2532331(direct) +31-30-2531492(secr.) Fax:
> > > +31-30-2518689
> > > > > > NIKHEF: +31-20-5922028(direct) +31-20-5922000(secr.) Fax:
> > > +31-20-5925155
> > > > > > CERN: +41-22-7679751(direct) +41-22-7675857(secr.) Fax:
> > > +41-22-7679480
> > > > > > Offices: Buys Ballot lab. 710 (Utrecht) H350 (Nikhef) B23 1-020
> > > (CERN)
> > > > >
> > > > --------------------------------------------------------------------------
> > > > > >
> > > > > >
> > > > > > Anton Fokin wrote:
> > > > > > >
> > > > > > > Hi,
> > > > > > >
> > > > > > > well, I was using Makefile included in the test directory. Yes, I
> > > saw
> > > > > > > Makefile.win32 in the same directory and I have tryied this one too
> > > (and
> > > > > it
> > > > > > > works somehow). Anyway, I suppose that if there is a Makefile in the
> > > > > binary
> > > > > > > distribution, it should be naturally related to this specific
> > > version of
> > > > > the
> > > > > > > distribution, so that a user can type nmake and get things compiled.
> > > > > > > Moreover, there is "win32" portion of lines in the Makefile, so that
> > > I
> > > > > > > assume this file is the correct one.
> > > > > > >
> > > > > > > Anyway, what I need is a makefile which works on both linux and
> > > win32,
> > > > > > > depending on a platform. Do we have such clever one? I am not
> > > win32/VC
> > > > > > > expert, so that I would like to get my linux root programs easily
> > > > > recompiled
> > > > > > > under win32. I thought I can use generic ROOT test Makefile upper
> > > part
> > > > > for
> > > > > > > that... Can I?
> > > > > > >
> > > > > > > Regards,
> > > > > > > Anton
> > > > > >
> > > > > >
> > > >
> > > >
>
> --
> Org: CERN, European Laboratory for Particle Physics.
> Mail: 1211 Geneve 23, Switzerland
> E-Mail: Fons.Rademakers@cern.ch Phone: +41 22 7679248
> WWW: http://root.cern.ch/~rdm/ Fax: +41 22 7677910
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:34 MET