Re: [Lprof-devel] Lprof QT4 Mac build

[Hal V. E. <hv...@as...>]

On Monday 21 January 2008 01:10:42 Mark Munte wrote:
> > The Qt4-port branch is pure devlopement and we expect to break
> > things.  Please
> > check it in and I will test and fix it on Linux.  I still need to
> > test this
> > on Windows as well.
>
> OK, it's all committed now.
>
> > Have you tried the cmake build?
>
> No, just the SCons build. Actually I never tried the cmake build. What
> is the future for Lprof, will the two build systems coexist or will
> you eventually drop one of them?
>

At this point I am not sure.  These can cooexist with each other but at some 
point I think we need to decide which way to go and drop the other one.   
Long term I do not want to maintain more than one build.

The cmake build does run faster since it has less startup overhead and scons 
being python is interpreted which makes it slightly slower.  But the scons 
build works with -j<x> where x is > 1 and the cmake build does not currently 
work if x is > 1.  So on my multi-core machine scons will do complete 
rebuilds faster than cmake even with the extra overhead but is slower for 
instances where a small number of modules are being rebuilt.

snip
> > Where are the help files located?  Are they being installed correctly?
>
> I compared the old (qt3) and new (qt4) Lprof. The folder structures
> are the same and all files are where they belong:
>
> lprof.app/MacOS
> 	assistant.app
> 	lprof (the actual binary)
> 	data (folder)
> 		help (folder, with 2 sub folders en, ru, html and other files inside)

OK this should work then.  So I don't know why this is happening.  

This code is located in src/lprofqt/config.cpp setupHelp(..).  And the section 
where we try to find the help files starts with:

if (help) delete help;
help = new QAssistantClient(execpath);  // so that help calls ...
if (found)
{
    // code of interest is located here
}

You need to check the value of path after the first set of if else if else 
statements in this block to see if it is pointing to the correct location.  
If it is not then we need to figure out why.  But first find out what value 
it has.  It should point to the lprof.app/MacOS/data/help/en directory in 
your case.  There are already commented out qDebug statements to look at 
this.

> 		pics (folder, files inside)
> 		profiles (folder files inside)
> 		template (folder, files inside)
> 		translations (folder, files inside)


> 		spyd2PLD.bin (this is where I copy the Spyder driver file so the
> Spyder gets recognized...)

No the spyder firmware belongs in the same directory as lprof not in the data 
directory.  There is a GUI now in LProf that will pop up when going to the 
calibration and measurement wizard that will allow you to install the 
firmware if it is not found.  This should install it in the correct location 
assuming you have write access to the installation directory.  I just tested 
this on linux with the Qt4-port branch and it worked.

>
> By the way, I tested the Spyder with this Lprof and it crashed when I
> tried to take readings for building the screen profile - same as with
> the old Lprof.
>
> > The OUTPUT_PROFILE value in the user configuration is set to whatever
> > the "Input Profile" combo box is set to when LProf is closed if you
> > have used
> > the Proof function.  If this is set to SPR200 Super... and this is
> > removed it
> > could cause a problem.  This is something that I don't think I have
> > tested
> > for so this could be an "undiscovered" bug in the code.  I just
> > tested this
> > on Linux (IE. renamed the file I had used in the Proof function) and
> > then
> > restarted LProf and it failed.  So this is a bug and I will get a
> > fix in soon
> > probably later today..

I got side tracked and have not located where this is happing yet.  I will try 
to get this fixed today.

>
> Yes, that was the problem: it was set to the SPR200 Super...icc
> profile. Changing that to some other profile solved the problem.
>
> Interestingly I now changed the "Input Profile" to Wide Gamut RGB,
> then closed Lprof, removed this profile and re-opened Lprof. Lprof now
> choose the next available profile and did not crash. So the bug
> happens under some circumstances only.
>

In my testing this would only happen if at some point you press the "Proof" 
button and then remove the profile that was used for the Proofing operation 
and then restart LProf.  When I locate where this is happening I will add a 
test to make sure the file exists() and isReadable()before making the lcms 
call that is causing this.

>
> I was able to "produce" another Lprof crash: under Camera/Scanner
> Profiler click the Select Directory button, then hit Cancel on the
> dialog displaying folders, then try do do something else like loading
> an image or selecting the reference file from the drop down: Lprof
> crashes.

This does not happen on Linux.  But I remember fixing something like this 
before and it is in the bug tracker so I will track down the bug report and 
see if I can find what I did to see if it might be related.  I will also test 
this on Windows when I get around to doing Windows testing. 

>
>
> Some other minor things I noted quickly going through the Lprof screens:
> 1) under Camera/Scanner Profile Parameters dialog -> general: the
> contents of the CLUT Resolution panel are very tightly spaced

I don't see this on Linux.

> 2) under Camera/Scanner Profile Parameters dialog -> White point
> handling: the Location panel is not rescaling with the screen and the
> contents are a bit mis-aligned

Yes I know.  This is a layout issue and I was not able to get this working the 
way I wanted it to and I had other things that were more critical so I moved 
on.  But it does need to be fixed.

> 3) under Camera/Scanner Profiler main screen: the Select Directory and
> Load Image buttons have a different look as all the other buttons:
> they are squared, all other buttons are rounded (the OS X default look)

I can see a slight difference on linux but it is mostly that these two are 
slightly lighter in color (so much so that it was easy to miss).  I will have 
a look at this.

> 4) under Monitor Profiler main screen: perhaps "Monitor" should be
> renamed to "Display" as some other screens use the word Display.... or
> the other way around...

Yes I agree.

> 5) under Display Profile Parameters dialog: the CLUT Resolution and
> Profile verbosity panels are very tightly spaced (as with point 1)

Again I am not seeing this on Linux.

> 6) under Monitor Profiler main screen -> build coarse profile -
> Monitor values dialog: the "OK", "Help" and "Set Gamma and Black
> Point" buttons are squared (as with 3)

Again on Linux this does not show up.  But the forms do need at least some 
work so I will have a look to see if I can figure out why these are different 
since it appears that some setting is likely different and this is affecting 
OS/X but not X11.

> 7) under Monitor Profiler main screen-> build accurate profile from
> measurements: both "Select Directory" buttons are squared (as with 3, 6)

These are created as part of the custom widgets in qtlcmswidgets.cpp.  I will 
look at the code to see if perhaps I need to change how these buttons are 
setup to make them more consistant with the rest of the interface.   This 
also applies to the "Select Directory" button issue in #3.

> 8) under Monitor Profiler main screen-> build accurate profile from
> measurements -> Calibrate and Measure: the dialog that opens has no
> title

The forms needed a lot of work after conversion from Qt3 to Qt4.  Things like 
layouts and the look of things was very messed up on some of the forms and I 
needed to touch almost everything.  But other forms did not need as much work 
and it appears that many of the above are for widgets that I did not touch as 
part of cleaning up the conversion since they looked OK on Linux.  So this 
will likely need more work to get everything looking good on all of our 
platforms.  For this stage of the conversion these are really minor issues 
since they don't affect actual functionality.  But I should have these fixed 
in the next few days.  

This is the kind of detailed feedback I was hoping for.  I should be able to 
have these sorted out fairly quickly and as these are fixed it will make the 
app look better.  So thanks. 

>
> Mark
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Lprof-devel mailing list
> Lpr...@li...
> https://lists.sourceforge.net/lists/listinfo/lprof-devel