Page 16 of 22

libltdl

Posted: Thu 15 Jan 2015, 16:06
by L18L

Code: Select all

# pdf2htmlEX 
pdf2htmlEX: error while loading shared libraries: libltdl.so.7: cannot open shared object file: No such file or directory
#

Posted: Thu 15 Jan 2015, 18:26
by SFR
@Neerajkolte: thanks, I'm glad you like it.
___________

Thank you L18L, the attached version works fine.
Although I've never been into translating and used it (momanager) to translate only a few of my own scripts, one day I must take a closer look at it. 8)

Thanks again &
Greetings!

Re: libltdl

Posted: Fri 16 Jan 2015, 07:19
by jamesbond
L18L wrote:

Code: Select all

# pdf2htmlEX 
pdf2htmlEX: error while loading shared libraries: libltdl.so.7: cannot open shared object file: No such file or directory
#
This, and the earlier problem with gphotofs that SFR pointed out much earlier, can be solved by installing libtool from the repo (or installing devx - it's in devx too). I don't know why the package depends on it (I will need to check their autoconf scripts) - libtool is a definitely a "dev" tool (together with compilers, autoconf, automake, etc). Anyway, if I can't resolve this in time, the simplest solution is to include libtool into the basesfs for the next release.

@SFR - About the aufs whiteouts. I have dug into snapmerge again and after a long thought (re-discovering the reasoning at the time I wrote the script), I come to the conclusion that with the dynamic loading/unloading of SFS, there is no way to guarantee correct behaviour with regards to whiteouts. There are two possible errors: 1) "un-deleted files" are deleted (ie hidden by the whiteout) - this is what you found out, 2) "deleted files" re-appearing again. I choose to ensure that 2) will never happen at the cost of 1) happening, because I think it is very unnerving to see a file that has been deleted showing up again. We may need to come for a simple GUI to help to clean-out all the whiteouts - but this has to be run consciously by the user, knowing the implications; rather than being run automatically at snapmerge or by load_sfs.

Samsung printer driver

Posted: Fri 16 Jan 2015, 14:35
by prehistoric
Just want to report that the samsung_print_fd600 driver version uld-1.00.6 in the 600 repository works with the Samsung M2625D I just acquired on both FD 630-631 and FD700 b2. Please note that this is not the same version as a previous Samsung unified driver you may have installed from the contributed repository.

I had previously had a failure in attempting to use this printer with Fatdog which I cannot now explain. Just wanted to tell people this is possible.

Re: libltdl

Posted: Fri 16 Jan 2015, 20:22
by L18L
jamesbond wrote:Anyway, if I can't resolve this in time, the simplest solution is to include libtool into the basesfs for the next release.
James,
thanks
a bit more fat is not bad
pdf2htmlEX is working now.
gslapt is super

@all do not forget to Update local package cache

Re: libltdl

Posted: Sat 17 Jan 2015, 00:08
by technosaurus
jamesbond wrote:
L18L wrote:

Code: Select all

# pdf2htmlEX 
pdf2htmlEX: error while loading shared libraries: libltdl.so.7: cannot open shared object file: No such file or directory
#
This, and the earlier problem with gphotofs that SFR pointed out much earlier, can be solved by installing libtool from the repo (or installing devx - it's in devx too). I don't know why the package depends on it
It probably doesn't actually depend on libltdl.so.7, but was linked in by mistake and without the --as-needed linker flag. You can either spend a bunch of time tracking down the random location that libtool/autotools decides to unnecessarily link it (its a royal PITA) in or just try rebuilding with -Wl,--as-neeeded added to the linker flags. ... a short term fix could be to make libltdl.so.7 a symlink libc.so (or whatever), however I don't know what order FD64 does layering for added sfs - the DEVX would need to go on a higher layer (it should always, but many puplets get it wrong, including some of Barry's official Puppy's)

Posted: Sun 18 Jan 2015, 04:12
by kirk
I think this is the reason of the issue:
Code:
#Debounce switch
[ -e /tmp/acpi_poweroff ] && exit
touch /tmp/acpi_poweroff #acpi_poweroff-delay.sh called later will delete this flag after a delay.

Later in the code is:
Code:
# don't suspend when lid is open
grep -q open /proc/acpi/button/lid/*/state && exit

so the acpi_poweroff flag remains in /tmp and blocks all further requests.
Thanks SFR, fixed. Well, the script is fixed, suspend not working is a whole other problem. Usually it's a bug in one of the kernel modules. If you can find out which one, and it's possible to remove it, then you can do that in the suspend script. If it's your video driver you're out of luck.

Sorry, been out for a while. Starting to get back into it.

momanager

Posted: Sun 18 Jan 2015, 13:22
by L18L
SFR wrote:Thank you L18L, the attached version works fine.
Although I've never been into translating and used it (momanager) to translate only a few of my own scripts, one day I must take a closer look at it. 8)
If you change

Code: Select all

        LANG=${ORIGLANG} xgettext -o ${ATEXTDOMAIN}.pot -j --from-code=UTF-8  --keyword=_ -L Shell --no-wrap  ./xEXTRASCRIPTS
to

Code: Select all

        LANG=${ORIGLANG} xgettext -o ${ATEXTDOMAIN}.pot -j --from-code=UTF-8  --keyword=_ -L Shell --no-wrap -f ./xEXTRASCRIPTS
(insert -f ) momanager will produce content of pot file and will need more time to do so 8)
My fault, sorry

Posted: Sun 18 Jan 2015, 18:30
by jamesbond
Page 14:
--------
@SFR: instead of xcompmgr you can try "compton" which is more versatile.
@SFR: "If I click one of the following entries in Razor's menu: Shutdown, Reboot, Suspend, Restart X, then cancel the action in confirmation dialog and click one of them again, very often the dialog won't appear. Have to click it 3rd time. " --- this is due to the "debounce" action related to the ACPI suspend stuff. You need to wait 2 seconds before re-doing any of them.

@SFR: "Trash" app is old. I'll take a look at your linked update, thanks.

@SFR: "It's ok if one's screen has 2048+ width and/or "Keep icons within screen limits" checkbox in ROX's Options is left in its default state (checked). " - I wasn't aware that Rox has that settings !!! I always thought that out of bound coordinates in Rox are always adjusted to fit into screen :) Might have to re-think that decision ...

@SFR: mplayer - I'm going to rebuild them. The liblive555 was removed because it was broken, but I forgot to update the dependency list. I'll do that later.

@SFR: gsmartcontrol desktop file needs some surgery, thanks.

@SFR: gphotofs: libltdl is part of libtool. In next release I'll have it in "libtool-libs" so you don't have to install the full libtool.
----

The story with the font is this: GTK apps depends on GTK to tell them the default font, Qt apps depends on Qt (which may depend on GTK) to tell them the default fonts, by raw X apps depends either on X resources to tell them the default font, or hardcode it altogether. After beta2 I have added a standard X resource (etc/X11/apps-default/Xresoures) that tells X-compliant apps to use DejaVu Sans as its default font - but as I said above, some apps don't respect this.

But most of the default fonts used by X apps are located in font-misc-misc package, so if you have this package installed, then you're great. It is *indeed* installed by default in the basesfs .... but only a subset of it, for size reasons :oops:. The problem is, our "choice" of what is "most commonly used" doesn't always correspond with the choice made by X apps authors (e.g in case of xosview).

There is another issue, which I think I've fixed in beta2, so that shouldn't matter anymore (in beta1 or earlier, "xlsfonts" may list fonts that don't exist).

In any case, installing font-misc-misc usually get things fixed.

----

I think I've reviewed the pages, and I have quite a long list of todo items. If I missed anything please let me/kirk know.

Re: libltdl

Posted: Sun 18 Jan 2015, 18:38
by jamesbond
technosaurus wrote:It probably doesn't actually depend on libltdl.so.7, but was linked in by mistake and without the --as-needed linker flag. You can either spend a bunch of time tracking down the random location that libtool/autotools decides to unnecessarily link it (its a royal PITA) in or just try rebuilding with -Wl,--as-neeeded added to the linker flags.
I'll give this a try, thanks.
... a short term fix could be to make libltdl.so.7 a symlink libc.so (or whatever), however I don't know what order FD64 does layering for added sfs - the DEVX would need to go on a higher layer (it should always, but many puplets get it wrong, including some of Barry's official Puppy's)
In Fatdog64 the basesfs is always at the lowest layer, all other SFS are higher than it (reverse of Puppy).

Posted: Mon 19 Jan 2015, 08:25
by rufwoof
@James

I see that FD 7b2 is using xz compression for remastering. A observation is that whilst that provides tight (smaller) output files there are less efficient alternative choices that produce larger output files, but process those files a lot quicker, for a overall net faster outcome.

I compiled LZO (lzop) into my Slacko 5.3.3 mksquashfs and after some analysis and reading around found that level 1 (lowest compression/fastest choice) was the best. Produced output file sizes around 20% larger than gzip, but processed around 3 times quicker, so around twice as fast overall. Interestingly since Kernel 3.11 LZ4 is included in the kernel (assuming its activated/compiled in), which decompresses around three times faster than LZO (around ten times quicker than gzip). The Kernel I'm currently running with however is 3.10 so I can't test that (and hence I've downloaded FD 7b2 to have a look at its Kernel 3.17).

The first thing I've done is remaster - and it takes a while. For my Slacko it takes 30 seconds or less on newer PC's, around a minute on older PC's, and I suspect that FD could see similar remaster times if it was using LZ4 (mksquashfs compiled with LZ4 support). Larger ISO, but loads - potentially much quicker.

I'm currently downloading FD DevX and Kernel Sources to have a look-see at the viability of adding LZ4 to mksquashfs. If that goes ok I might try using that in fatdog-remaster to see what differences become apparent.

Great work with FD, your talent and effort are admirable.

Posted: Mon 19 Jan 2015, 12:25
by SFR
jamesbond wrote:@SFR: instead of xcompmgr you can try "compton" which is more versatile.
Yeah, already did, because xcompmgr has a bug - from time to time (dunno what's triggers this, though) quite a large, top-left part of the screen becomes kinda frozen and unusable.
Compton is much better.
jamesbond wrote:If I missed anything please let me/kirk know.
Please note that L18L's mod of xlock gui has fixed "only" the crash in case of missing translations; there are still LANG=en_US.UTF-8 (*), CTRL+ALT+Backspace, VT switching, '-grabserver' & lack of "Reset" button issues:
http://www.murga-linux.com/puppy/viewto ... 032#816032

Btw, dclock screensaver fixed itself after installing those additional X fonts.

(*) What's interesting, some other .UTF-8 LANGs, like en_GB.UTF-8 or en_CA.UTF-8 don't show this problem - only en_US.UTF-8.
Also, just noticed (not sure if it's supposed to be like that) it's impossible to change Localisation in Control Panel from en_US to any of the other en_?? options:
"Cancelled, nothing is changed".
___________

bcrypt_gui has problem with some special characters.
Here's fixed version (the one for older Puppies):
http://www.murga-linux.com/puppy/viewto ... 735#788735
___________

filemnt

Code: Select all

mountpoint=$(echo $fullpath | tr "\\/ .;:?*[]{}\t" +)
An apostrophe (') also needs to be replaced with +.
___________

[later]
jamesbond wrote:- thanks for the updated swf.xml - I'll use it.
I forgot that it's also necessary to:

Code: Select all

cd /usr/share/ROX-Filer/ROX/MIME/
ln -s application-x-shockwave-flash.png application-vnd.adobe.flash.movie.png
cd /etc/xdg/rox.sourceforge.net/MIME-types/
ln -s application_x-shockwave-flash application_vnd.adobe.flash.movie
___________

[even later]
PeasyGlue doesn't work due to trimmed down netpbm stuff (lack of jpegtopnm, pnmtojpeg, etc.).
PeasyScale works partially (only 'transformation' mode) due to lack of peasyscale.bin.
PeasyPrint won't work without GS.

Greetings!

English locales

Posted: Mon 19 Jan 2015, 14:16
by L18L
SFR wrote:(*) What's interesting, some other .UTF-8 LANGs, like en_GB.UTF-8 or en_CA.UTF-8 don't show this problem - only en_US.UTF-8.
Also, just noticed (not sure if it's supposed to be like that) it's impossible to change Localisation in Control Panel from en_US to any of the other en_?? options:
"Cancelled, nothing is changed".
Confirm "Cancelled, nothing is changed" if NLS.sfs not loaded (savefile=none).

After having loaded NLS.sfs I tried to add English for Denmark, dialog says OK but:

Code: Select all

# cat $HOME/.fatdog/language
en_DK.UTF-8
# 
# echo $LANG
en_DK.UTF-8
# 
# locale -a
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
C
POSIX
en_AU
en_AU.utf8
en_US
en_US.utf8
# 
Strange :roll:


EDIT
/usr/lib64/locale/en_DK is an empty directory
Hope that helps

Posted: Tue 20 Jan 2015, 09:13
by olinm
Video nasty: Two big bugs in VLC media player's core library

http://www.theregister.co.uk/2015/01/20 ... xec_flaws/

Flaws disclosed in late December await exploitation

Posted: Tue 20 Jan 2015, 22:32
by technosaurus
SFR wrote:PeasyGlue doesn't work due to trimmed down netpbm stuff (lack of jpegtopnm, pnmtojpeg, etc.).
PeasyScale works partially (only 'transformation' mode) due to lack of peasyscale.bin.
PeasyPrint won't work without GS.

Greetings!
I once patched the netpbm package to work as a multicall binary (like busybox) that significantly reduced the installed size. I've been meaning to submit a patch.... regarding the P*app dependencies, please file a bug report with the upstream Puppy developer to properly specify dependencies in the package

A couple other things I noticed

I forgot to mention that dpkg and dpkg-deb are missing directory and file in /var ... I think their binaries/symlinks are also in the wrong directory as well ... came across it when I was trying to unpack Chrome as an AppDir.

Chrome also reported missing libexif (which is also needed by fottox and other image utilities)

I also didn't see a way to install FD64 to disk after booting from a unetbootin installed FD64 usb drive ... at least not in the control panel or start menu. I always run from usb anyhow, but realize others may be looking to install to disk.

Also, since FD64 doesn't have a first run wizard, you may have to manually set date and time to get https to work on computers that don't have time/date set properly already (for ex. new computers that haven't been run yet, or old ones with dead CMOS battery)

I mentioned earlier my issues with my intel sound card (where device 1 is the actual device) and posted an alsa conf entry to fix it... something like that could probably be added to a sound wizard.

The jwm package is missing the .jwmrc-tray (possibly others)

Posted: Tue 20 Jan 2015, 22:32
by technosaurus
double post

Posted: Wed 21 Jan 2015, 06:40
by jamesbond
@SFR:
- bcrypt_gui - will update, thanks [EDIT] - ends up with a total re-write.
- filemnt - will update, thanks [EDIT] fixed.
- Rox MIME icons - will update, thanks. [EDIT] fixed.
- peasyglue problem: netpbm was improperly built, it's fixed now (=re-install netpbm from repo)
- peasyscale problem: peasyscale.bin was renamed to peasyscale-cli, but I forgot to update the script. Will fix, thanks. [EDIT] fixed, new package has peasyscale.bin symlinked to peasyscale-cli.
- peasyprint won't work without GS - yes, this is intended. You should get a message saying peasyprint won't work if GS is not installed.

@L18L:
- missing locale settings ... that's odd. I'll look into it.

@olinm:
- Will have to wait until VLC team releases next release

@technosaurus:
- dpkg and dpkg-deb are missing directory and file in /var ... I think their binaries/symlinks are also in the wrong directory as well ... came across it when I was trying to unpack Chrome as an AppDir
[EDIT] dpkg package (from the repo) seems to have all the correct directories. The built-in dpkg/dpkg-deb is from busybox is of course missing them; they are meant only for "unpacking" .debs and not for proper installation.

- Chrome also reported missing libexif (which is also needed by fottox and other image utilities) - [EDIT] libexif is already in the repo.

- I also didn't see a way to install FD64 to disk after booting from a unetbootin installed FD64 usb drive ... at least not in the control panel or start menu. - I will have to try this myself, to see how unetbootin put the files into USB, then perhaps update the installer.

- Also, since FD64 doesn't have a first run wizard, you may have to manually set date and time to get https to work on computers that don't have time/date set properly already (for ex. new computers that haven't been run yet, or old ones with dead CMOS battery) - you can enable the "ntpd-client" service to sync your clock from internet time servers at boot time. It won't work for first boot, but after you have a savefile it would work.

- I mentioned earlier my issues with my intel sound card (where device 1 is the actual device) and posted an alsa conf entry to fix it... something like that could probably be added to a sound wizard. --> sound wizard only works with ALSA devices (so far). And there is no relationship between ALSA devices and OSS (/dev/dsp*) devices. For example, in my laptop, I have two ALSA devices (hw:0 for analog output and hw:1 for HDMI output) but there is only one /dev/dsp. Please print the content of your /proc/asound/oss/devices and /proc/asound/oss/sndstat for comparison?

We'll see if the problem is common enough to add management of OSS devices into sound wizard. Alternatively, you can always do the symlink yourself in rc.local and/or edit /etc/udev/rules.d/54-fatdog.rules and symlink /dev/dsp to /dev/dsp1 instead.

- The jwm package is missing the .jwmrc-tray (possibly others) - [EDIT] jwm package doesn't contain pre-configured configs. The .jwmrc-tray and other config is part of fatdog base configuration.

[EDIT]
@rufwoof: for my own use I'm the basesfs is "gzip -1" compressed. It is larger but since I'm running with 8GB of RAM, I can live with that. But we're using the xz compression for general distribution so people with only 1-2GB RAM can use it well enough. The mksquashfs tool in Fatdog only supports gzip and xz, I haven't done the research to get it to support LZO/LZ4 (it should be able to, since the kernel squashfs module does support both).

EDIT: update some responses.

first run

Posted: Wed 21 Jan 2015, 09:10
by L18L
technosaurus wrote:Also, since FD64 doesn't have a first run wizard, you may have to manually set date and time to get https to work on computers that don't have time/date set properly already (for ex. new computers that haven't been run yet, or old ones with dead CMOS battery)
I have always been stumbling upon false keyboard layout before
Correct time is essential for some basic functions.
Assuming English only users need a locale too I suggest:

Fatdog's first run that is if $HOME/.fatdog/language is unset
could be

Code: Select all

choose language
force launch of the 4 dialogs in Localization of Control Panel (no cancel )
:wink:

Re: first run

Posted: Thu 22 Jan 2015, 06:33
by jamesbond
L18L wrote:

Code: Select all

choose language
force launch of the 4 dialogs in Localization of Control Panel (no cancel )
:wink:
Haha okay, how about this instead, I'll just a (self-destruct) desktop icon with title "First-time localisation setup" or something like that? It's self-destruct, after you complete the localisation then the desktop icon will disappear.

Re: English locales

Posted: Thu 22 Jan 2015, 17:37
by jamesbond
L18L wrote:
SFR wrote:(*) What's interesting, some other .UTF-8 LANGs, like en_GB.UTF-8 or en_CA.UTF-8 don't show this problem - only en_US.UTF-8.
Also, just noticed (not sure if it's supposed to be like that) it's impossible to change Localisation in Control Panel from en_US to any of the other en_?? options:
"Cancelled, nothing is changed".
Confirm "Cancelled, nothing is changed" if NLS.sfs not loaded (savefile=none).

After having loaded NLS.sfs I tried to add English for Denmark, dialog says OK but:

Code: Select all

# cat $HOME/.fatdog/language
en_DK.UTF-8
# 
# echo $LANG
en_DK.UTF-8
# 
# locale -a
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
C
POSIX
en_AU
en_AU.utf8
en_US
en_US.utf8
# 
Strange :roll:


EDIT
/usr/lib64/locale/en_DK is an empty directory
Hope that helps
Found the problem. This is a build system bug. Some required files in /usr/share/i18n/locales are missing, even to generate en_* locales (these files are moved to nls.sfs). Fixed. (as a note, those files are iso*t1 files and i18n file. Certain locale e.g. en_DK actually needs da_DK also).