[paludis-user] cave print-unused-distfiles vs. binary repos

[denisgolovan]

Hi all

Recently, I came across a strange "cave print-unused-distfiles" behaviour.
It treats _installed_ packages from binary repos as unused.

Just like following:
denis at den-desktop ~ $ cave show gcc-win-x32
* sys-devel/gcc-win-x32
::installed 4.6.3 {:4.6}
::local-bin 4.6.3 {:4.6}
::overlay 4.6.3* {:4.6}
sys-devel/gcc-win-x32-4.6.3:4.6::installed
Description The GNU Compiler Collection
Homepage http://gcc.gnu.org/
From repositories local-bin overlay
==================skipped==================================

Even though my custom gcc build was definitely installed using binary repo "local-bin", "cave print-unused-distfiles" decided it unused.
"cave print-unused-distfiles" is a part of my weekly unused distfiles deletion, those binaries were just removed.

I've read a manual for "print-unused-distfiles" and I think it's an error.
Please fix it.

[denisgolovan]

Hi again

Sorry, I forgot to mention:

denis at den-desktop ~ $ cave --version
paludis 1.4.0

[Ciaran McCreesh]

On Mon, 24 Jun 2013 21:51:19 +0400

Post by denisgolovan
Recently, I came across a strange "cave print-unused-distfiles"
behaviour. It treats _installed_ packages from binary repos as unused.
denis at den-desktop ~ $ cave show gcc-win-x32
* sys-devel/gcc-win-x32
::installed 4.6.3 {:4.6}
::local-bin 4.6.3 {:4.6}
::overlay 4.6.3* {:4.6}
sys-devel/gcc-win-x32-4.6.3:4.6::installed
Description The GNU Compiler Collection
Homepage http://gcc.gnu.org/
From repositories local-bin overlay
==================skipped==================================
Even though my custom gcc build was definitely installed using binary
repo "local-bin", "cave print-unused-distfiles" decided it unused.
"cave print-unused-distfiles" is a part of my weekly unused distfiles
deletion, those binaries were just removed.
I've read a manual for "print-unused-distfiles"

-i , --include
Treat all distfiles from IDs in the specified repository as
used. May be specified multiple times. Typically this is used
for binary repositories, to avoid treating non-installed binary
distfiles as unused.

[denisgolovan]

???-i , --include
???????Treat all distfiles from IDs in the specified repository as
???????used. May be specified multiple times. Typically this is used
???????for binary repositories, to avoid treating non-installed binary
???????distfiles as unused.

Hm. I read it like "use to non-installed packages".
Why it should be used for installed ones?

[Ciaran McCreesh]

On Mon, 24 Jun 2013 22:03:30 +0400

Post by denisgolovan

???-i , --include
???????Treat all distfiles from IDs in the specified repository as
???????used. May be specified multiple times. Typically this is used
???????for binary repositories, to avoid treating non-installed
binary distfiles as unused.

Hm. I read it like "use to non-installed packages".
Why it should be used for installed ones?

It's down to how we track where an ID is fetched from. Once an ID is
installed, fetches_key comes from the VDB / exndbam entry from the
original package, not the binary. We mostly do this to avoid presenting
strange metadata.

Possibly there should be a second key there tracking binary downloads
that's also considered for 'used' distfiles.

[denisgolovan]

Post by Ciaran McCreesh
It's down to how we track where an ID is fetched from. Once an ID is
installed, fetches_key comes from the VDB / exndbam entry from the
original package, not the binary. We mostly do this to avoid presenting
strange metadata.
Possibly there should be a second key there tracking binary downloads
that's also considered for 'used' distfiles.

Aha!
So, it's a bug :).

[Ciaran McCreesh]

On Mon, 24 Jun 2013 23:21:07 +0400

Post by denisgolovan

Post by Ciaran McCreesh
It's down to how we track where an ID is fetched from. Once an ID is
installed, fetches_key comes from the VDB / exndbam entry from the
original package, not the binary. We mostly do this to avoid
presenting strange metadata.
Possibly there should be a second key there tracking binary
downloads that's also considered for 'used' distfiles.

Aha!
So, it's a bug :).

No, it's doing exactly what it's supposed to do! The question is merely
whether what it is supposed to do is the same as what users suppose it
does, and if not then whose suppositions need to be altered.