site  contact  subhomenews

GTK "gseal" fix

December 09, 2011 — BarryK
Wary and Racy 5.2.2 users have reported applications failing to compile, with "gseal" error messages.

I had gone through a rather tiring process of getting GTK to work properly prior to releasing 5.2.2, and didn't feel like getting into that again. Here are some of the pre-5.2.2 deliberations:

Well, the "gseal" problem needed to be fixed. I have rolled back to GTK 2.24.5, compiled with this:

# ./configure --prefix=/usr --sysconfdir=/etc --localstatedir=/var --build=i486-t2-linux-gnu --with-xinput=yes --enable-debug=minimum --enable-introspection=no

After installing it, I tested 'quicksetup' and the dropdown lists are fast (that was a big problem before). I compiled gtktetris, that ttuuxxx had reported gave the "gseal" errors -- now it compiles.

Here are the PETs (2.3MB, 266KB):

The next releases of Wary and Racy will be built with these.


Midori crashes

December 01, 2011 — BarryK
I reported earlier that I was unable to compile Webkit, needed for Midori web browser:

Jemimah replied that where it was failing is a utility application that can be removed from the makefile. Yes indeed, I did that, and it compiled. Then I compiled Midori 0.4.2.

At startup, Midori partly loaded the Google home page, then crashed.

Weird, Midori is loading libraries inside /initrd/pup_rw/usr/lib/ -- that is so incredibly weird, as that is definitely not in the library search path. It must be something peculiar that the 'waf' build system does.

Jemimah has also compiled Midori 0.4.2, for Racy 5.2.2: tester so far, reports it crashes.

So much for Midori.


Gtkterm rs232 serial terminal

November 20, 2011 — BarryK
I posted about the Cutecom GUI serial terminal in the previous blog post:

There is a later version of Cutecom, that requires Qt4:

There is also a GTK-based GUI serial terminal, named Gtkterm. I discovered that it is still being developed. It needs VTE. Site:

I tried to compile these apps statically against Qt4 and VTE, but failed. Probably it could be figured out. In the latter case there were undefined symbols, but applying what I thought were logical steps to fix it, did not. So, left it and just compiled Gtkterm again shared VTE library. Here are the PETs:

Gtkterm (37KB):

VTE (356KB):


Wary Puppy 5.2.2

November 18, 2011 — BarryK
I am pleased to announce the release of Wary Puppy version 5.2.2, our build of Puppy Linux that targets older hardware. This is a minor upgrade and bug-fix release of Wary 5.2. Though, if you look in the release notes you will might not think it so "minor".
This is also the debut of Racy Puppy, which we think of as "Wary on steroids". Racy enhances Wary to run on recent hardware, with Xorg 7.6 and 3.0.7 kernel.

Full announcement and release notes:

Download Wary (130.5MB):

Download Racy (111.6MB):

Forum thread for Wary feedback:

Forum thread for Racy feedback:


Wary and Racy 5.2.2 "TESTING"

November 18, 2011 — BarryK
If anyone is interested in conducting a "sanity check", here they are:

About several hours from now I intend to remove the "TESTING" from the directory names and formally announce the releases.


'welcome1stboot' fixed

November 17, 2011 — BarryK
A few days ago ...14th?... I had Wary and Racy at what I thought was release quality. Then I found that 'welcome1stboot' would not start after I had upgraded GTK from 2.24.5 to 2.24.8.

(On the other-hand, other testers reported that welcome1stboot did work.)

So, I played with rolling GTK back to 2.20.1.

But, just now I have discovered why welcome1stboot does not work. I am both relieved and very exasperated.

'welcome1stboot' is written in BaCon. Now, BaCon does not link the GTK libraries as shared libraries in the normal manner. Instead, BaCon has it's own code to load the libraries. This is where the problem occurs -- and it sure is awful, I consider it to be a serious bug in BaCon.

Wary/Racy has these in /usr/lib, where the first is a symlink:

When the 'devx' is loaded, there is also the symlink:

Now this is the awful part: at runtime, the Bacon app tries to load ''. It cannot, so crashes. There is no error message identifying the cause of the crash.

It is standard, engraved in stone, practice that *.so files are used during compiling, not at runtime. The BaCon application should really be using

God, I have spent days working on this problem, on the verge of "developer fatigue". Oh well. I will put the *.so symlinks into the gtk PET, and hopefully should be able to put up another Release Candidate of Wary/Racy.


Wary (and Racy)

November 15, 2011 — BarryK
I tried an experiment. Although it was the "wrong" thing to do, I built Wary/Racy with the GTK 2.20.1 PETs from Wary 5.1.x. The reason that it is "wrong" is that the core packages in 5.2.x were recompiled in T2, with GTK 2.24.5 and some later libraries (ex: pixman, cairo, glib). However, as reported, there are problems with GTK 2.24.x. To roll GTK back by just plunking in the PETs from Wary 5.1.x is just asking for trouble.

Or so it would seem. I built it, and everything seems to work. Well, I haven't checked printing, that is one area where there might be some difference between the GTK versions.

Terryphi posted about "tester fatigue". Yes, I am getting "developer fatigue" with these troubles with GTK. I really want to get Wary and Racy 5.2.2 released, then move on to other stuff -- multi-architecture Woof and ARM Puppy. So, give a run, if it looks ok, I'll go with that for the final.

I will be off-line for approximately the next 48 hours, maybe slightly less. Starting from about now. Racy has been uploaded, Wary is still uploading, eta 15 minutes from now:



As soon as the upload has completed, I will be going off-line. Got other family-related stuff to do for a couple of days.

Please go here to provide Wary feedback:

Please go here for Racy feedback:


GTK trouble

November 14, 2011 — BarryK
See my previous post "GTK success!":

Yeah, well, now the pendulum has swung the other way. I built Wary and Racy, even uploaded them. However, after doing a frugal install, I found that 'welcome1stboot' segfaults.

I recompiled it in Wary, it then worked. I used that new executable in Racy ....aaaargh, it still segfaulted!

I regret moving up from GTK 2.20.1. With that version the ROX-Filer focus problem that came in at GTK 2.18.x, was partly fixed. Well, with GTK 2.24.x that "part fix" is gone again.

I wonder if it would break anything if I did roll back to GTK 2.20.1? The problem is, I did a complete compile of all the core packages in T2, against GTK 2.24.5. Rolling GTK back would be asking for trouble.