Core v4.5rc1

[roberts]

That group of 4 work on the currently selected extension.

the first 3 of these 4 items are perform the other functions
might be better to separate this item also from upper 3 items?

One can argue either way, should it be group by function or group by required selection.

[roberts]

1. With "Load local", I am using a standard FLTK widget that just so happens to offer "favorites" where you can create a shortcut to selecting a file. I my drop that standard widget and just reuse some of my existing code. That way, I could display only "not already loaded" extensions that would be more practical and consistent with the rest of the layout.

2. I presently do not have plans to merge the two. My reasoning, is that it leaves open the possibility to have a tcz only system, or even an scm only system. Brian, has updated his ezremaster to support scms. So by placing the "extended base" tczs into the Core.gz one could easily remaster an scm only system.  Besides, it was enough of a challenge to merge AppBrowser with AppsAudit with all the permutations and combinations of GUI elements.

Re: Load Local
I am dropping the standard FLTK file selector widget and replace with my code.
The benefits are:
1. Display only available, "not already loaded" extensions. Should provide a much shorter select list and therefore make the system easier to use and understand.
2 Consistent look, feel, and operation, i.e., everything is single click and will work like ScmApps Local/Install.
3. The generic file selector is heavy with its preview and favorites, both not needed, thereby its removal will mean a small memory footprint.

Done!

An additional benefit is that one does not need OnDemand, at least it will empower those using window managers and desktop environments that do not support Ondemand an easy and clear way to select from a list of locally available unloaded extensions.

[Juanito]

The mirror is shown at bottom of GUI, was moved from title bar.

..must be going blind - sorry 

[roberts]

I forced some md5 errors while performing updates. Found and fixed an anomaly.
The Apps GUI now properly report failures as well as successes on the result of md5sum checking with the Updates function.

[AbNoRMiS]

i have already reported here about missing checksum
cde/optional/wireless-3.0.21-tinycore.tcz.md5.txt
in the CorePlus-4.xx.iso
and about what can remove empty cde/optional/libiconv.tcz.dep
although in principle it does not interfere anyone

[roberts]

CorePlus items updated. Thanks!

[SvOlli]

Please re-test the pxe/tftp support:

Since the the desktop is loaded as an extension, I get an error message when booting up with a different user. I looks like the loading and installing is done using the provided user account, but the account will only be created after all tczs are loaded.

Greetings,
SvOlli

[curaga]

@SvOlli

see http://forum.tinycorelinux.net/index.php?topic=12722.0

[SvOlli]

@SvOlli

see http://forum.tinycorelinux.net/index.php?topic=12722.0

Thanks for the hint, but this is only part of the problem: when I specify a user at kernel command line, booting via tftp breaks completely, because an "su $USER -c ..." is done with no existing user, so su fails before handling the tcz. And this did work prior to 4.2.

What would break if the code segment of /etc/init.d/tc-config

Code: [Select]

if [ -n "$USER" ]; then
        if ! grep "$USER" /etc/passwd >/dev/null; then addUser; fi
else
        USER="tc"
fi
is moved before

Code: [Select]

# First process tftp and http entries
modprobe -q squashfs 2>/dev/null
[ -n "$TFTPLIST" ] && get_app "TFTP" $TFTPLIST
[ -n "$HTTPLIST" ] && get_app "HTTP" $HTTPLIST
?

It would be a fix for the problem I mentioned.

[gerald_clark]

tftp/http loading was originally written to load nfs drivers for tce directory on nfs.
It was then extended to support nbd and aoe.
Since these are system rather than user extensions user tc was used.

You want to use it to load user extensions, and that patch looks like it will work and make tftp/http loading more useful.

I suggest you test it, and submit a patch to roberts.

[SvOlli]

tftp/http loading was originally written to load nfs drivers for tce directory on nfs.
It was then extended to support nbd and aoe.
Since these are system rather than user extensions user tc was used.

Sorry, but this explanation has nothing to do with the bug I reported, since loading nfs, nbd and aoe via tftp/http also breaks if you pass a user= option to the kernel commandline:

From /etc/init.d/tc-config:

Code: [Select]

get_app() {
[...code removed...]
su "$USER" -c "/usr/bin/tce-load -i $FL $TCE/$APP"
[...code removed...]
}

# First process tftp and http entries
modprobe -q squashfs 2>/dev/null
[ -n "$TFTPLIST" ] && get_app "TFTP" $TFTPLIST
[ -n "$HTTPLIST" ] && get_app "HTTP" $HTTPLIST

[...code removed...]

if [ -n "$USER" ]; then
if ! grep "$USER" /etc/passwd >/dev/null; then addUser; fi
else
USER="tc"
fi

So, first the nfs module gets loaded using "su "$USER" -c "..."" and after that that same user will be generated. This has to fail.

This is definitely a bug which wasn't in there in 4.1, and I'm reporting it, handing it over on a silver plate. As a workaround I dropped the user= kernel parameter, which made my system usable again. Sorry, but I've been out of the loop for about half a year now, and I just don't have the time to fix and test it, since I'm buried beneath other projects.

BTW: I still wonder why package installation is only allowed for the user and not for root, imho it should be exactly the other way round, as in any other distribution I've encountered so far.

What happend to the old days when, all you got for reporting a bug was: "Fixed! Thanks for finding.", and not an endless discussion?

[gerald_clark]

The explanation has everything to do with the bug you reported.
Testing is simple.
Create a second initrd that contains only tc-config.
Add that to the initrd line in your pxeboot config file.

Boot and verify it is working.

You are the one wishing to extend the original purpose of the boot code.

[SvOlli]

You are the one wishing to extend the original purpose of the boot code.

You are wrong here. Do not wish extend any original purpose. If you've got a pxe/tftp/http setup, please just add the option "user=gerald" to the kernel options, and it not a single TCZ will load. As I stated:

As a workaround I dropped the user= kernel parameter, which made my system usable again.

So it works for me, still I just wanted to report that problem.

And trying to make myself as clear as possible:
- I found a bug.
- I reported it.
- The bug is triggered in a well defined and completely valid environment, one that is not on the "not supported"-list.
- I don't want any extension of any feature.
- I don't even want this to be fixed, I just wanted to report it to see if anyone feels responsible for it.
- Since there no-one seems to be, I've got no problem with it being in there up to version 4.99 or whatever.
- And I'm not willing to sink more time into this problem, as I've got a workaround, and projects that are more fun to hack on as TCL.

Yes the time spent in discussion might have been enough to fix the problem already. But this goes for both sides, and I didn't see this coming when I filed in my bug report.

[gerald_clark]

I'm just a user, like you, but I made the change, tested it and sent a patch to roberts.

[roberts]

@SvOlli, Thanks for the bug report and suggested fix.

@Gerald, Thanks for making the effort to test and submitting a patch. Greatly appreciated!

Look for this beginning with 4.5rc2.