Merge ~ahasenack/ubuntu/+source/squid:cosmic-squid-4.2-merge into ubuntu/+source/squid:debian/sid
Status: | Work in progress |
---|---|
Proposed branch: | ~ahasenack/ubuntu/+source/squid:cosmic-squid-4.2-merge |
Merge into: | ubuntu/+source/squid:debian/sid |
Diff against target: |
417 lines (+298/-4) 11 files modified
debian/NEWS (+0/-0) debian/changelog (+113/-0) debian/control (+5/-3) debian/patches/90-cf.data.ubuntu.patch (+16/-0) debian/patches/99-ubuntu-ssl-cert-snakeoil.patch (+22/-0) debian/patches/fix-uninitialized-var.patch (+25/-0) debian/patches/series (+3/-0) debian/rules (+11/-1) debian/squid.install (+3/-0) debian/squid.preinst (+15/-0) debian/usr.sbin.squid (+85/-0) |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Robie Basak | Approve | ||
Canonical Server | Pending | ||
Review via email: mp+356100@code.launchpad.net |
Description of the change
Bileto ticket: https:/
PPA: https:/
DEP8 results: https:/
Merge from debian's 4.2-2 in an attempt to fix #1794553, where upstream commented that the crash might be due to incomplete gcc8 fixes that are in 4.1 and that were improved in 4.2.
There is no evidence yet that 4.2-2 fixes the reported crash, though. Also, upstream's 4.3 has further gcc8 fixes. I'm also using squid from cosmic in my home proxy and haven't seen it crash yet.
This MP is to have a review in place, and address any issues it raises, so that if the reporter of the crash bug can eventually confirm that the crash is gone, we can land this branch without further delays.
I had to introduce one small delta to correct the last debian change which was to add -latomic to certain architectures. The way it was done, it was being added to all arches, not just the listed ones. I pushed a PR to salsa to fix that.
I spent some time investigating this debian change: "Add upstream pr264 patch for systemd (Closes: #903165)". It claims to also resolve launchpad bug #1103362, but I'm not sure. I can confirm that squid fails to start if it cannot resolve a hostname in an acl, but I failed to reproduce that case in a normal boot of the system, even when bind9 was installed and configured (via netplan) to be the sole dns resolver of this system.
Also of note about that fix is the fact that what was eventually merged upstream did not include the "Wants=
debian's patch: https:/
Upstream commit: https:/
The debian patch is now removed in salsa for the upcoming 4.3 release, exactly because upstream adopted it, but sans the "Wants=" line.
I can adopt the exact same change that upstream did, or keep debian's from 4.2-2. Right now this MP has debian's patch. As I said, I couldn't reproduce the mentioned bug (https:/
About the remaining gcc8 fixes, upstream hasn't committed all of them yet, so I kept my two changes in that regard: d/p/fix-
Unmerged commits
- bb82787... by Andreas Hasenack
- 8538d8f... by Andreas Hasenack
- 9b5b0c6... by Andreas Hasenack
- 1190dd5... by Andreas Hasenack
- 1b9f272... by Andreas Hasenack
- 9e2270c... by Andreas Hasenack
- 75cd25d... by Andreas Hasenack
- 5573dde... by Andreas Hasenack
- de21835... by Andreas Hasenack
- ec27139... by Andreas Hasenack
lgtm
The use of empty commits in the final branch to mark the Dropped changes for use by reconstruct- changelog are interfering with range-diff. Something to talk about with Christian perhaps from a workflow perspective. But we can discuss that later.