[pkg-fso-maint] An eventful install.sh

Steffen Möller steffen_moeller at gmx.de
Fri Mar 5 21:57:13 UTC 2010


Dear Steven,

I have added your patch and just nibbled at a few corners. For instance
the internal variable names are prepended with "internal_" now. I thank
you for indicating "dpkg -s" to me - have not been aware of it.

Steven De Herdt wrote:
> All right,
> 
> I was not aware of those demands. A test on the memory would probably
>> complicate the script too much. I'll add a respective warning to "--help",
>> is that fine?
>>
> I would agree that this has the best ROI, we can expect the user to RTFineM.

Uh, I had forgotten about this one. Added.

> Hm. This should possibly be another flag. Luca (IIRC) objected adding less
>> critical packages (also like apt-utils) to keep the install truly minimal.
>> For us "SD-Installers" this is less of an issue, but when aiming at the
>> flash, you don't want any bloat.
>>
> Something like SD_LUXURY that pulls these packages in then.  Maybe suppres
> those perl warnings if not set? (emacs says I got 133 of those, ten lines
> each)

I admit to be not overly much interested. In my humble opinion the bug
is more with the tool the utters all the redundant warnings than with us
poor souls using it.

The tasks are meant to address the individuals' diverse foci of
interest. A task "luxury" would not be sufficiently expressive and ...
well ... I did not suffer that much about it so far (puh!)

> Bash is not my mother tongue (...), but I tried to translate those comments
> the best I could.  I started from commit
> 09fb172d5044c977474cfa0c7f5fff320bb5dab4, with which the attached patch
> should match.  It is wholly divided in parts three, of which:
> *one changes the debootstrap version to one that does inhabit the
> repositories;
> *another makes the cleanup stage only purge installed packages (may contain
> historic cruft);
> *the third hopefully clarifies the umount message and corrects the mount
> instructions.  Those instructions are only shown when !
> $intern_instdir_already_present, is it assumed that the presence of that
> directory signals that the current OS already handles the mounting of the SD
> card?  SHR mounts mmcblk0p1 on /media/card for me...
> These parts actually don't differ between each other concerning language.

It look all fine to me. Concerning the DELENDA, er, internal_delenda, I
wished we would kind of determine the packages to delete when we know if
we are using (c)?debootstrap ..... or find some other more deterministic
approach. Or some flag to dpkg that allows the specification of
non-existing packages for removal? Well, since your code certainly fixes
something, I am happy for how it is for the moment.

> As for the erratic SD card, I suppose I'll hunt for a solution separately.
> It seems to be a rather complex matter.

Please just something smallish as a start. As a sidenote, my internet
provider (Alice) just freed me from GPRS (and UMTS, if I had it)
charges. This gives online storage a completely new meaning to me. So,
go small, keep your documents online :)    And, should anyone change to
Alice because of this notice, ask me for my phone number and tell them,
I'll donate the 50€ reward as a travel grant to those pkg-fso developers
who want them.

> P.S.: If this silly humour in mail or patch are bothersome or impertinent I
> do apologise.

Gosh. I have not even spotted it. I presume, one just needs it for what
we are doing here.

Cheers,

Steffen




More information about the pkg-fso-maint mailing list