4.3 Beta 1 -- bugs and reports

[Sit Heel Speak]

***edited: bug in microscope corrected, it now sees Fedora 11 correctly, and distinguishes "not present" from "is not set"***

***EDITED again: but, see my next post. b43, I now believe, should be not set, period. Don't put it in the kernel (or compile as a module) at all.***

By using the dotconfig-comparison utility which I describe in
http://www.murga-linux.com/puppy/viewtopic.php?t=45831
I see the following, with regard to B43 and B44 dotconfig switches. By looking at this screenshot you can see which distros' dotconfigs I am comparing.

(Suggestions as to most-advanced distro's whose dotconfigs I should be examining under this "12-way blink-comparator microscope" are welcome--I am not yet enough of a Linux guru to know what the 12 best are. The program runs in a DOS window under Windows 98SE. Memory limitations of this combination, mainly the upper limit on open file handles, determine that I can compare a maximum of only 12 distro's at a time.)

Not present in vanilla 2.6.29.6 nor foreirongold's,
not set in all the others:
*# CONFIG_B43LEGACY_DMA_MODE is not set
*# CONFIG_B43LEGACY_PIO_MODE is not set
*# CONFIG_B43_DEBUG is not set


Set in all, except is not set in vanilla 2.6.29.6 and foreirongold's:
*CONFIG_B43=m
*CONFIG_B43LEGACY=m
*CONFIG_B44=m

Set in all, except is not present in vanilla 2.6.29.6 and foreirongold's:
*CONFIG_B43LEGACY_DMA=y
*CONFIG_B43LEGACY_DMA_AND_PIO_MODE=y
*CONFIG_B43LEGACY_PCICORE_AUTOSELECT=y
*CONFIG_B43LEGACY_PCI_AUTOSELECT=y
*CONFIG_B43LEGACY_PIO=y
*CONFIG_B43_PCICORE_AUTOSELECT=y
*CONFIG_B43_PCI_AUTOSELECT=y
*CONFIG_B44_PCI=y
*CONFIG_B44_PCICORE_AUTOSELECT=y
*CONFIG_B44_PCI_AUTOSELECT=y

Set in Ubuntu 9.04 and CrunchBang Linux (Ubuntu 9.04.1 with Openbox),
not present in vanilla 2.6.29.6 and foreirongold's,
is not set in any of the others:
*CONFIG_B43LEGACY_DEBUG=y

Set in the Pup(pie)s and Fedora 11 but not set in Ubuntu 9.04 and CrunchBang;
not present in vanilla 2.6.29.6 and foreirongold's:
*CONFIG_B43_PCMCIA=y

Set in Fedora 11, not present in vanilla 2.6.29.6, Ubuntu 9.04,
Crunchbang, Puppy 4.12-k2.6.25.16 nor foreirongold's,
set in all other Pup(pie)s:
*CONFIG_B43_PIO=y

Set in all except not present in vanilla 2.6.29.6, Puppy 4.12 2.6.25.16, MU's config-2.6.29.1MU-based-on2.6.28.5, and foreirongold's:
*CONFIG_B43LEGACY_RFKILL=y
*CONFIG_B43_RFKILL=y

Set in all, except is not present in unpatched vanilla 2.6.29.6, Puppy 4.12 2.6.25.16, and foreirongold's:
*CONFIG_B43LEGACY_LEDS is set =y
*CONFIG_B43_LEDS=y

Sorry for the messup in the previous version. There was a fencepost error in the parser, such that the first finding of a present-and-set switch when stepping through the distro's (as often as not, this was in Fedora) was not being recorded. Alles in ordnung now.

[jaddison]

First boot on lenovo r61 and MSI Wind:

No wireless on either. MSI not a surprise; had to add 8187 pet with 4.21. Lenovo a little surprising--on 4.21 no wireless when I run the "connect" wizard, but if I use the Pwireless app to scan for a connection and then manually connect I usually have no problem. No dice either way with this version.

4.3 gives me sound on the MSI machine--never could get that to work with 4.21.

[capoverde]

No problems found up to now after one week with Puppy 4.3Beta1 (single-proc), real fine.

After reading Barry's blog post about Wine I downloaded wine-1.1.28-i486.pet and installed it, then ran the Win$ version of a music notation program - MuseScore, mscore-0.9.4.exe - which runs perfectly and way faster than the Linux version (actually MuseScore 0.9.3 on Puppy 4.1)!

It sees all fonts by putting a symlink to the fonts directory, which is in another ext3 partition, into Wine's "Windows-> Fonts" directory -- this 6-7yr-old PC has no Win$.
How funny, it's just the opposite of what one needs to do in order to have TrueType fonts in Puppy with Win$ boxes! I'm quite amazed, Wine's performance is far better than I expected.

[paulh177]

frugal install

I tried a .pup conversion only to find the following :

Code: Select all

# pup2pet xxx.pet
.pup packages are no longer supported in Puppy,
This conversion script is no longer supported either.
#

Searching hasn't revealed a statement from BK or anyone else about this, but I may have overlooked it.
.

[Sit Heel Speak]

I have now *actually read* tempestuous's thread from last November, "Extra drivers for Puppy 4.1 with 2.6.25.16 kernel" --his message with a hybrid Broadcom wifi driver PET attached on page 5...and, followed up by *actually looking at* what Broadcom themselves say, at
http://www.broadcom.com/docs/linux_sta/README.txt
Broadcom recommends you use their hybrid driver. This readme.txt paragraph seems particularly relevant:

On this or a machine with the same kernel version, install the driver.

1. Validate you don't have loaded (or built into the kernel) the Linux community provided
driver for Broadcom hardware. This exists in two forms: either "bcm43xx" or a split form
of "b43" plus "b43legacy". If these modules were loaded you would either
a) rmmod bcm43xx or
b) rmmod b43; rmmod b43legacy

So it seems my post, above, is misleading, in that it implies that some combination of b43 switches at compile-time might be the golden key...when, in reality, you don't want any Linux-community-made b43 drivers at all, instead you want the Broadcom driver. Adding b43 etc. switches at compile time, probably messes things up.

[fratermus]

Hi All
I was just about to say I have a bug then read the post about incompatible sfs's. But did not think that was going to affect the save files. At one time I think I remember puppylinux offering to upgrade a save file or something like that. If I was to rename it to the new name system would it then offer to convert it.

Watching with interest. As a puppy newb I am confused about how one will be able to use their 4.2x pupsave file in the 4.3 version.

4.3 doesn't seem to see the 4.21's pup_save_crypta.2fs on the hard drive. Does this mean we will effectively have to clean install 4.3 when it becomes a release, and reconfigure by hand?

It's not a dealbreaker, I just don't think I understand what is supposed to happen.

[rerwin]

Although I have sent this information to Barry via PM, the latest message has remained in my outbox since 8/25, so I am re-posting it here, that it not get lost.

I followed directions in the ubuntu community wiki for merging part of the oem-version of the HSF modem driver into the latest version from Linuxant, to retain the best of both. I added a step that allows it to be compiled for either of our kernels (and probably any others). I have documented all that I did to create the source tarball, compile and configure it, and update the firmware tarball. I also documented the compiling, configuring and firmware updating for the HCF and DGC modem drivers.

Attached are the HSF-full-dell-hybrid source tarball and the instructions intended for addition to Barry's "kernel.modules.txt" file, which explains how to compile and integrate each special driver. I have already sent Barry the HCF and DGC source tarballs. I hope all three will be added to the puppylinux.com/sources collection for both kernels.

The resultant (k2.6.29.6) dotpet for all three drivers is here:http://www.murga-linux.com/puppy/viewto ... 081#335081
Richard

[capoverde]

Seeking in the thread I didn't find a mention about this: testing Puppy 4.3 Beta1 on a friend's PC, I accidentally saw that when the mouse pointer stays over a drive icon, the filesystem description comes out OK but the capacity is always 0,0GB.
Didn't mind about that on my machine; will have a look when back home.

* * * Edited: Sure, happens on my box too.
...And the <Insert> key bug is back... Awwww...

[rerwin]

Attached is the Linuxant 1.0.20 patch for application to the alsa-driver-1.0.20 source package. It supports "high definition audio" HSF modems. I have applied the patch and compiled the alsa-driver successfully. I cannot attach the resultant source tarball as it is almost 4 MB and is rejected by the forum attacher.

I have no way of testing it, because (1) I don't have such a modem and (2) the beta1 on which I built it is still at level 1.0.16. I can provide the dotpet once beta2 is released, assuming beta2 uses alsa-driver-1.0.20.

Here is my "kernel.modules.txt" entry for both patching and building it:

alsa-driver-1.0.20-hda-patched
To patch an extracted original alsa-driver, obtain the linuxant patch matching the alsa-driver version, to directory containing the extracted alsa-driver directory (home), then:
# cd
Make Linuxant patch:
# patch -p0 < alsa-driver-1.0.20-2.patch
Rename as patched driver, for Conexant HSF HDA modems:
# mv alsa-driver-1.0.20 alsa-driver-1.0.20-hda-patched
Make patched source tarball:
# tar -czf alsa-driver-1.0.20-hda-patched.tar.gz alsa-driver-1.0.20-hda-patched
End of patched-source tarball creation
Compile the alsa driver:
# cd alsa-driver-1.0.20-hda-patched
# ./configure --with-cards=hda-codec-conexant,hda-codec,hda-intel
# make
# new2dir make install
...manual install alsa-driver-1.0.20-hda-patched-i486/lib/modules/.../kernel/sound/* to kernel/sound

Although there is mention on the Web of a patch for Agere HDA modems, that patch appears to be unnecessary with 1.0.20, as I get errors if I try it.
Richard

[BarryK]

Barry,

Thanks for the response

1. Pcur is great, but default folder should not be in the list also, same as Rox
2. b43 issue, I have to get back to you tomorrow, I don't have the computer with b43 with me now (it's 8.30 PM now here )

regards

Trio,
The script writes to a log file, /root/test_b43.log.gz

[BarryK]

Regarding the problem with the firmware not installing for b43 module, sorry but my previous test script has a bug.

Trio and others, would you be so kind as to test with this script:

[BarryK]

...probably best to test with "pfix=ram" and of course be sure that your hardware that needs the b43 module is present.

[sinc]

i believe the first line in the terminal said something to the effect of 'b43 module not found in /proc/modules' but i use the b43legacy module in this computer (i'm posting from it in 2.14x8 right now). heres the log from the script. let me know if you want me to try something else. thanks.

this was from a clean install with no pupsave.

[trio]

Barry,

Here's the log you asked (with pfix=ram). Also please notice the jpg I attached.

Tq

[James C]

I've been testing 423 Beta (2.6.25.16 and 2.6.29.6 kernels) live saving to usb stick. Tonight I decided to do a full hard drive install.....no success.

The installer identified my ext3 partition as ext4, proceeded anyway .Completed installation, rebooted and......nothing. Booted into a different Puppy on a different partition and checked where I installed 430 Beta,or 423 in grub, and discovered a totally empty partition. No files were copied during the installation.

Reformated and tried again, same negative results.Any ideas?

[rerwin]

I have compiled the latest version of the non-HDA drivers for Agere/agrsm modems, patched to resolve a compiler error and using the 2.6.30.5 kernel source. Here are the "kernel.modules.txt" instructions for patching and building; the patched source tarball; and a dotpet containing the 2 agr modules, updated udev rules to support all modems specified in the driver source, and an updated firmware tarball where the script now allows for use of an optional HDA version of the drivers.

The HDA option will apply only to the puppy version using kernel 2.6.25.16, because the HDA driver cannot be compiled for later kernels due a GPL issue regarding proprietary code. But I want to avoid multiple versions of the script. (My plan is to provide an HDA version of the drivers when the standard version of beta2 is available, as an optional dotpet, to be conservative.)

The instructions:

agrsm-20090502-patched
To patch the original tarball, extract agrsm-20090502.tar.gz, which creates directory agrsm-20090502, then:
# cd agrsm-20090502
Patch use of low-level function call (to use higher-level interface same as other drivers):
# sed -i -e 's/__request_region(&ioport_resource, /request_region(/' agrsoftmodem.c
Create patched tarball:
# cd
# tar -czf agrsm-20090502-patched.tar.gz agrsm-20090502
End of patching process. Start here if using patched tarball.
# cd agrsm-20090502
# make
# make install
The warnings about "section mismatches" can be ignored.
...manual install /lib/modules/.../extra/agr* to extra

Richard

[playdayz]

Code: Select all

sh-3.00# ./iron
./iron: /usr/lib/libgcc_s.so.1: version `GCC_4.0.0' not found (required by ./iron)
./iron: /lib/libc.so.6: version `GLIBC_2.7' not found (required by ./iron)
./iron: /lib/libc.so.6: version `GLIBC_2.8' not found (required by ./iron)
./iron: /usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.9' not found (required by ./iron)
./iron: /usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.10' not found (required by ./iron)
sh-3.00# 

This is what happens when I try to run SRWare Iron alpha for linux (the spyware free version of google chrome). It is not a bug in 4.3 of course but I am wondering if 4.3 will be able to run Iron or should be able to run Iron or could be able to run Iron or would be able to run Iron with some hacks....

It (4.3) will be able to run Seamonkey 2.0 which is noticeably faster than the 1.1.x versions of Seamonkey (on benchmarks and in normal use), and Iron is noticeably faster again--it will finally give Linux a browser that is equal to the fastest browsers available anywhere. And if not in 4.3 then in 5 I hope. I spend an enormous amount of time working in browsers and the speed is a welcome usability feature. thanks.

[Tui]

Att: Mr. Barry K, I hope you read my post.
Please revert to the ALL in your next Puppy beta version,
I would like to see it back in the PETget package manager as I find it better to
go thro' A - Z when looking for a package as I don't know what packages are
filled under.

Thanks
Tui

[BarryK]

I posted this notice on my blog, but in case anyone didn't read it...

4.3beta2 should be out in a day or two, so if you are interested in testing a 4.3 beta then hold off a bit. Lots of bugs in 4.3beta1 have been fixed.

[rerwin]

After many false starts I have patched and compiled the driver for Intel 536 modems. However, I have not be able to test on my modem. When attempting to compile for 2.6.29.6 I get some errors that I could try to work around and then test, but I submit my results so far for either inclusion or user testing on beta2. The dotpet has the driver file and the updated firmware tarball. I do not want to delay submission any longer.
UPDATE: The dotpet attachment is not quite correct: wrong /lib/modules/<kernel>". The corrected dotpet is now in the beta2 forum, here: http://www.murga-linux.com/puppy/viewto ... 644#338644

Here is my "kernel.modules.txt" entry for the Intel536:

intel-536EP-2.56.76.0_2009_02_08-patched
Create patched version for all kernels, by extracting tarball intel-536EP-2.56.76.0_2009_02_08, which creates directory intel-536EP-2.56.76.0, then:
# cd intel-536EP-2.56.76.0
Patch to avoid using member 'owner', removed from kernel 2.6.30:
# sed -i -e 's%\( *\)\(hamproc->owner = THIS_MODULE;\)%\1#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,30)\n\1\2\n\1#endif%' coredrv/coredrv.c
Create patched tarball:
# cd
# tar -czf intel-536EP-2.56.76.0_2009_02_08-patched.tar.gz intel-536EP-2.56.76.0
End of patching process. Start here if using patched tarball.
# cd intel-536EP-2.56.76.0
# make clean
# make 536
# make install
The make-install will end with an error message that can be ignored, as the error occurs on the final command of the script:
unknown distribution - no boot scripts have been installed
make: *** [install] Error 1
...manual install /lib/modules/.../kernel/drivers/char/Intel536.ko to extra
[NOTE THAT THE DRIVER PREVIOUSLY WAS IN /kernel/drivers/char/, SO THIS WOULD BE A CHANGE TOWARD CONSISTENCY WITH OTHER PUPPY-MAINTAINED DRIVERS -- IF BARRY CONCURS.]
...also userland stuff - Update the intel536ep firmware tarball:
- copy /usr/sbin/hamregistry into usr/sbin/ in the /lib/modules/all-firmware/intel536ep tarball.

The dotpet contains the updated firmware tarball. I have also compiled this for kernel 2.6.25.16 and it tests successfully. (I understand this may be too late for the beta2, but the dotpet can be installed by testers.)

BTW, the 537 driver appears to be a more difficult task, so will take some time.
Richard