On Thu, 4 Oct 2012, Jon Trulson wrote:

> It will take a it for it to propagate around the planet, but it's up
> there.  Here's the relevant, and somewhat brief portion of the HISTORY
> file:

Congratulations to everyone!

Here is a write up of my work in progress as well
as some thoughts on the nearest future:

tl;dr:
 - outstanding patches
 - complaints re XmPrivate.h
 - thoughts on libcsa interop
 - thoughts on versioning



83a60bb Add -Wno-write-strings to C++ compiler flags
Posted 24 Sep 2012 to
http://thread.gmane.org/gmane.comp.desktop.cde.devel/614

  I understand that we want to have a nicer set of
  compiler specific options in config/cf to be able
  to set warnings set per compiler.
  
  On the principle of -Wno-write-strings I find the
  mailing list thread inconclusive :)

  Here is how FreeBSD introduces various levels
  of warnings for two compilers supported:

  http://svnweb.freebsd.org/base/head/share/mk/bsd.sys.mk?revision=240468

  So the code grows from being WARNS=1 to WARNS=6 ready
  (WARNS=6 being -Wall and much more).

  Maybe we could have something similar and apply
  different "warning readiness" levels for various
  components (it will be difficult to do it for
  a whole tree). So something like

    WARNS = 3 

  could be added to those Imakefiles for components
  we already improved.

  Any thoughts?

296b1fb FreeBSD 10: Don't include <wchar.h> in dtksh
Post 27 Sep 2012 to
http://thread.gmane.org/gmane.comp.desktop.cde.devel/669

  I don't know if our understanding is any better now.
  I am using this patch all the time because otherwise
  it won't build on FreeBSD 10.x I am now running.

eb22182 Adobe Helvetica as -dt-interface system-
Discussed: http://thread.gmane.org/gmane.comp.desktop.cde.devel/656/focus=683

  Unpublished patch changing default font for
  -dt-interface system- from
    "-b&h-lucida-medium-r-normal-sans-*" to
    "-adobe-helvetica-medium-r-normal--*".
  I think there was a slight agreement on this 
  mailing list that we can have it as the new
  default (at least for FreeBSD). Linux uses
  serif -b&h-lucidabright-* for this, maybe
  you would prefer a sans serif font too?

13fcf5e dtcm: 64bit work in progress
   
  Unpublished work in progress to fix some
  64-bit issues in dtcm. Unfortunately, I think
  I took a slightly wrong approach there
  and I need to review those changes again:

  Just changing 
    
     Dimension x; XtVaGetValues(..., MnSomeX, &x, ...);
  
  to

     XtArgVal x; XtVaGetValues(..., MnSomeX, &x, ...);

  doesn't always work as expected if some calculations
  are done on "x" later.

  In some cases I actually have to do
   
     XtArgVal x_;
     Dimension x;
     XtVaGetValues(..., MnSomeX, &x_, ...);
     x = (Dimension)x_;

  for example here:

diff --git a/cde/programs/dtcm/dtcm/misc.c b/cde/programs/dtcm/dtcm/misc.c
index 1a9bc2f..8168f08 100644
--- a/cde/programs/dtcm/dtcm/misc.c
+++ b/cde/programs/dtcm/dtcm/misc.c
@@ -1238,6 +1238,8 @@ ComputeMaxWidth(
 {
        Dimension       width1,
                        width2;
+       XtArgVal        width1_,
+                       width2_;
        Widget          w[5];
        int             i;
 
@@ -1250,16 +1252,18 @@ ComputeMaxWidth(
        w[4] = (Widget) NULL;
 
        XtVaGetValues(w1,
-               XmNwidth,               &width1,
+               XmNwidth,               &width1_,
                NULL);
+       width1 = (int)width1_;
 
        for (i = 1; w[i]; i++) {
 
                XtVaGetValues(w[i],
-                       XmNwidth,       &width2,
+                       XmNwidth,       &width2_,
                        NULL);
+               width2 = (int)width2_;
 
-               if ((int)width2 > (int)width1) width1 = width2;
+               if (width2 > width1) width1 = width2;
        }
 
        return width1;

   I am using this patch now but it's probably not
   ready for the master.

----

* XmPrivate.h situation

I am also *very* frustrated with the XmPrivate.h
as it undermines my basic understanding of how
make works. Maybe there is some problem with
sub-second timestamping, I'm not sure. I would
be grateful for any help to full understand
how git and make get it right on my platform
and wrong on Jon's.

----

I have two crashes outstanding:

- crash of dtcreate upon exit due to double free();

#6  0x0000000000417e95 in free_Filetypedata (pFtD=0x8059e2320) at af_aux.c:570
570       if (pFtD->pszOpenCmd) XtFree(pFtD->pszOpenCmd);

<jemalloc>: jemalloc_arena.c:173: Failed assertion: "bitmap_get(bitmap, &bin_inf
o->bitmap_info, regind)"

looks like we are freeing something we have already
freeed.

- dtfile crash:

#0  0x000000000043d89f in UpdateHeaders (file_mgr_rec=0x5a5a5a5a5a5a5a5a, 
    file_mgr_data=0x805cf9940, icons_changed=1 '\001') at FileMgr.c:2899
2899       if ((file_mgr_data->show_iconic_path == 0) !=

don't know how to reproduce, happens seldomly and is not
very intrusive. Something heap-related too.
  
----

I have also abandoned some changes to libcsa because
it seems tricky to get it right. It seems like
that right now in my build (CSA_uint32) is long (64-bit)
and not uint32_t despite what the name says. 
Trying to force it into 32 bits seems not a good idea
either but I *hope* we are still interperable because
those values got serialized into XDR "long" type.

Anyway we need to check whether 64-bit dtcm can
connect to 32-bit rpc.cmsd and vice versa, any
volunteers?

-----

I am going to look into how DtSvc detects 
CDE runtime version vs. compiled version
because of this message:

_DtXlateGetXlateEnv: Platform: FreeBSD; Execution Ver: 0; Compiled Ver: 10000

I think we could add something like __CDE_version
to tell everyone the version of CDE running.

Maybe even we could patch "dthello" or introduce
"dtversion" utility to tell the release and
git commit ID of the version running - I've
seen already people reporting problems against
2.2.0a or 2.2.0b release that are already fixed
in master.

As we may soon run out of alphabet letters for
our releases, what about introducing a Solaris-like
scheme like CDE 2.2.0 2012/10 release? We could
have also __CDE_version set to 20121001 or something
like that. CDE-2.2.0-20121001 would also fit
nicely into most packaging schemes until we are
brave enough to bump it to 2.2.1
  
//Marcin

------------------------------------------------------------------------------
Don't let slow site performance ruin your business. Deploy New Relic APM
Deploy New Relic app performance management and know exactly
what is happening inside your Ruby, Python, PHP, Java, and .NET app
Try New Relic at no cost today and get our sweet Data Nerd shirt too!
http://p.sf.net/sfu/newrelic-dev2dev
_______________________________________________
cdesktopenv-devel mailing list
cdesktopenv-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/cdesktopenv-devel

Reply via email to