Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 581166 - net-misc/curl - Please provide sub slot
Summary: net-misc/curl - Please provide sub slot
Status: RESOLVED UPSTREAM
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: Current packages (show other bugs)
Hardware: All Linux
: Normal enhancement (vote)
Assignee: Anthony Basile
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-04-25 12:42 UTC by Thomas Deutschmann (RETIRED)
Modified: 2016-05-10 15:24 UTC (History)
1 user (show)

See Also:
Package list:
Runtime testing required: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Thomas Deutschmann (RETIRED) gentoo-dev 2016-04-25 12:42:48 UTC
Please consider providing a sub slot (for current net-misc/curl-7.48.0 I would suggest SLOT="0/4.4.0", SONAME version of libcurl.so) which would allow ebuilds which link against libcurl to set a slot operator on net-misc/curl so that an update of net-misc/curl can trigger a rebuild for these applications when necessary.
Comment 1 Anthony Basile gentoo-dev 2016-04-28 00:28:19 UTC
(In reply to Thomas Deutschmann from comment #0)
> Please consider providing a sub slot (for current net-misc/curl-7.48.0 I
> would suggest SLOT="0/4.4.0", SONAME version of libcurl.so) which would
> allow ebuilds which link against libcurl to set a slot operator on
> net-misc/curl so that an update of net-misc/curl can trigger a rebuild for
> these applications when necessary.

so curl has this funny option:

  --enable-soname-bump    Enable enforced SONAME bump
  --disable-soname-bump   Disable enforced SONAME bump

which triggers different behaviour in lib/Makefile which is commented.  i'll repeat the comment here but you should read the code for the full story:

# Bumping of SONAME conditionally may seem like a weird thing to do, and yeah
# it is. The problem is that we try to avoid the bump as hard as possible, but
# yet it is still necessary for a few rare situations. The configure script will
# attempt to figure out these situations, and it can be forced to consider this
# to be such a case! See README.curl_off_t for further details.

Currently the ebuild has --disable-soname-bump.  So I'm not against adding a subslot, but if I do, then we need to switch to --enable-soname-bump.  But following that comment, this should be "necessary for a few rare situations", so my question is, why do we need this in gentoo?
Comment 2 Anthony Basile gentoo-dev 2016-05-07 18:02:41 UTC
@Thomas, ping.  can you give me more context as to why you want this in light of comment #1?
Comment 3 Thomas Deutschmann (RETIRED) gentoo-dev 2016-05-09 15:59:15 UTC
I don't have an actual reason. I filled this request after reviewing multiple ebuilds which are all depending on net-misc/curl.

Following the ongoing process of providing/using sub slots I thought net-misc/curl should also provide one to allow emerge to trigger an automatic rebuild when necessary for these packages.

However like you have shown net-misc/curl seems to be different and the project claims to do everything they can to prevent breakage of backward compatibility by design.

So if you agree we can close this as "Won't fix". Ebuilds depending on net-misc/curl can already depend on net-misc/curl:0= if they want to.
If we will ever encounter a breakage we can still decide to set a sub slot which then will trigger a rebuild for those ebuilds which already subscribed to net-misc/curl's not yet explicit set sub slot.
Comment 4 Anthony Basile gentoo-dev 2016-05-10 15:24:23 UTC
(In reply to Thomas Deutschmann from comment #3)
> I don't have an actual reason. I filled this request after reviewing
> multiple ebuilds which are all depending on net-misc/curl.
> 
> Following the ongoing process of providing/using sub slots I thought
> net-misc/curl should also provide one to allow emerge to trigger an
> automatic rebuild when necessary for these packages.
> 
> However like you have shown net-misc/curl seems to be different and the
> project claims to do everything they can to prevent breakage of backward
> compatibility by design.
> 
> So if you agree we can close this as "Won't fix". Ebuilds depending on
> net-misc/curl can already depend on net-misc/curl:0= if they want to.
> If we will ever encounter a breakage we can still decide to set a sub slot
> which then will trigger a rebuild for those ebuilds which already subscribed
> to net-misc/curl's not yet explicit set sub slot.

triggering rebuilds is a necessary evil.  if we have to in order to avoid breakage, we do.  but curl upstream claims we can avoid it, so we will.

its resolved upstream, sorta ;)