Re: Control file
Justin B Rye <justin.byam.rye@gmail.com> writes:
> Finally a patch. Extra notes:
>
> * I've squeezed the first paragraph, and (usually) merged the last two.
Wow - i think I almost understand what these packages are for now. I
also thought the technical info helped a lot, and was really interesting
without being overwhelming.
In the first para "in terms of package dependencies" could perhaps be
clearer as "using/via/though package dependencies"? - or maybe "allow
ISA requirements to be represented as package dependencies"?
The bit i dont understand is why the use of pre-depends is correct:
eg,
- suppose i dont have sse3 support in my computer
- i can open aptitude and tell it i want to install gnome
- gnome requires the virtual www-browser which chromium provdes
- I can tell the package manager to use chromium to satisfy the dep
- I can now start the installaton of hundreds of packages (which will
include sse3-support) - all standard stuff to here.
- sse3-support will eventually fail to install - the pre-depends means
this happens before chromium, but i only learn this part-way through a
multiple-megabyte download of hundreds of gnome-related packages: seems
to me i might be left with a bunch of cryptic dpkg errors and not much
to hint that this is due to chromium and not gnome. (i doubt many people
would remember that i manually asked for chromium)
.... wouldnt it be simpler if chromium _conflicted_ with a package that
represented "this computer does not support sse3"? then i couldnt even
start installing chromium unless the cpu was up to it?
.... wouldnt it be even better for chromium to detect the missing sse3 at
runtime, not at package install time
... what happens if i change my cpu to add/remove sse3 support but keep
the same disk and same installation? (if it's a raspberry pi i can
easily move the disk between cpus)
(i also also note that in stable, chromium only depends on sse3-support,
not pre-depends. maybe that changed in sid - but i thought chromium was
always backported)
(i also suspect a name ending "-support" was perhaps not the most
helpful one from the point of view of a user coming accross these in a
package list - sounds like a set of drivers/firmware to me - but i can
see that it makes writing "Depends: sse3-support" read nice for
developers.)
R
Reply to: