Project

General

Profile

Actions
Copy link

Bug #10691

closed

ceph-dencoder links to libtcmalloc, and shouldn't

Added by Ken Dreyer about 9 years ago. Updated about 8 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

Source:
other
Tags:
Backport:
hammer,firefly
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

The /usr/bin/ceph-dencoder binary links to libtcmalloc, and it has no need to link to it.

From a packaging perspective, since /usr/bin/ceph-dencoder ships in the "ceph-common" RPM, this means that ceph-common requires the gperftools-libs RPM to be installed. Ideally we'd avoid this dependency for ceph-common.

Related issues 2 (0 open — 2 closed)

Copied to Ceph - Backport #12461: ceph-dencoder links to libtcmalloc, and shouldn'tRejectedBoris RantoActions
Copied to Ceph - Backport #12511: ceph-dencoder links to libtcmalloc, and shouldn'tResolved01/29/2015Actions
Actions
Copy link
 #1

Updated by Greg Farnum about 9 years ago

  • Status changed from New to 12

Looks like the linkage is sneaking in via the autotools LIBMDS variable (which itself includes LIBPERFGLUE). We don't actually use LIBMDS very often, and most of the places also don't want tcmalloc, so we can probably just drop LIBPERFGLUE from LIBMDS and include it explicitly in the MDS build.

Actions
Copy link
 #2

Updated by Ken Dreyer about 9 years ago

Thanks Greg, I think you're right. Yesterday evening when I talked with Dan about this, he was musing whether it would make sense to split LIBMDS to LIBMDS_TYPES, similar to what was done in 1c170776cb8c0266f0f54f049ed07bbdb9c9ab5e ? The commit log is a little terse there but we were thinking that commit was done in order to reduce ceph-dencoder linkage.

Actions
Copy link
 #3

Updated by Greg Farnum about 9 years ago

I'd love to have a LIBMDS_TYPES, but the reason we didn't do so already is that it's much harder than for the other systems. The MDS base types, their in-memory representations, and their code are a lot more tangled together than in the other systems. We're closer than we were back then but I wouldn't try out that avenue. :(

Actions
Copy link
 #4

Updated by Ken Dreyer about 9 years ago

  • Status changed from 12 to Fix Under Review
Actions
Copy link
 #5

Updated by Sage Weil about 9 years ago

  • Status changed from Fix Under Review to Won't Fix

fixed downstream. if we want to resolve this later we can use their approach (two version of libmds.la basically)

Actions
Copy link
 #6

Updated by Ken Dreyer about 9 years ago

@Sage Weil, this is still an issue - could we keep this tracker open?

Actions
Copy link
 #7

Updated by Ken Dreyer about 9 years ago

  • Status changed from Won't Fix to Fix Under Review
Actions
Copy link
 #8

Updated by Sage Weil about 9 years ago

  • Status changed from Fix Under Review to Resolved
Actions
Copy link
 #9

Updated by Ken Dreyer about 9 years ago

The final fix for this was https://github.com/ceph/ceph/pull/4359 . Thanks Boris!

Actions
Copy link
 #10

Updated by Ken Dreyer almost 9 years ago

  • Status changed from Resolved to Pending Backport
  • Backport set to hammer
Actions
Copy link
 #12

Updated by Loïc Dachary almost 9 years ago

  • Status changed from Pending Backport to Resolved
  • Regression set to No
Actions
Copy link
 #13

Updated by Loïc Dachary almost 9 years ago

  • Status changed from Resolved to Pending Backport
  • Backport changed from hammer to hammer,firefly
Actions
Copy link
 #14

Updated by Loïc Dachary about 8 years ago

  • Status changed from Pending Backport to Resolved
Actions
Copy link

Also available in: Atom PDF