Bumping the compiler version to 6.2.0 breaks 64bit Tor Browser builds

After bumping the GCC version to 6.2.0 due to #13893 (moved) building GCC 6.2.0 is busted for 64bits in our gitian setup:

/usr/bin/ld.bfd.real: gcov.o: relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC

added GeorgKoppen201610 TorBrowserTeam201803R boklm201803 component::applications/tor browser owner::tbb-team parent::25304 priority::very high resolution::duplicate severity::blocker status::closed tbb-rbm type::defect labels

This works fine with hardening-wrapper settings as we have them in Gitian on a Debian testing machine. Additionally, setting DEB_BUILD_HARDENING_PIE=0 for the GCC step in our Gitian build would help as well.

The Wheezy compiler is too old for it (generally 4.x).

Okay, I bisected the problem (PIE enforcement breaks the compiler) and pushed a fix for review to bug_20301 (https://gitweb.torproject.org/user/gk/tor-browser-bundle.git/commit/?h=bug_20301&id=dfa3403825fc315168df3295f6062f5979b0a8a0) in my public repo.

Trac:
Keywords: TorBrowserTeam201610 deleted, GeorgKoppen201610, TorBrowserTeam201610R added
Status: new to needs_review

Taking off the review queue as GCC 6 seems to be too unstable to use right now. See #20381 (moved) for further details.

Trac:
Keywords: TorBrowserTeam201610R deleted, TorBrowserTeam201610 added
Status: needs_review to new

Moving over to rbm

Trac:
Keywords: tbb-gitian deleted, tbb-rbm added

Trac:
Keywords: N/A deleted, boklm201802 added

Trac:
Keywords: TorBrowserTeam201610 deleted, TorBrowserTeam201802 added

Any reason you went with 6.2? I'm using 6.4 (for 32 bit) in TaskCluster.

Replying to tom:

Any reason you went with 6.2? I'm using 6.4 (for 32 bit) in TaskCluster.

I think when this ticket was opened (October 2016), 6.2 was the last version. We should probably use 6.4 now.

Replying to boklm:

I think when this ticket was opened (October 2016), 6.2 was the last version. We should probably use 6.4 now. 7.3 now.

Replying to boklm:

Replying to tom:

Any reason you went with 6.2? I'm using 6.4 (for 32 bit) in TaskCluster.

I think when this ticket was opened (October 2016), 6.2 was the last version. We should probably use 6.4 now.

Yes, exactly.

Trac:
Parent: N/A to #25304 (moved)

boklm201802 -> boklm201803

Trac:
Keywords: boklm201802 deleted, boklm201803 added

Adding to our March plate.

Trac:
Keywords: TorBrowserTeam201802 deleted, TorBrowserTeam201803 added

Replying to gk:

Okay, I bisected the problem (PIE enforcement breaks the compiler) and pushed a fix for review to bug_20301 (https://gitweb.torproject.org/user/gk/tor-browser-bundle.git/commit/?h=bug_20301&id=dfa3403825fc315168df3295f6062f5979b0a8a0) in my public repo.

I applied the same fix, using the same comment, to tor-browser-build in branch bug_20301: https://gitweb.torproject.org/user/boklm/tor-browser-build.git/commit/?h=bug_20301&id=4e57f8a0780fd9ed9633dc7d2391bf908b74e337

Trac:
Status: new to needs_review
Keywords: TorBrowserTeam201803 deleted, TorBrowserTeam201803R added

Replying to boklm:

Replying to gk:

Okay, I bisected the problem (PIE enforcement breaks the compiler) and pushed a fix for review to bug_20301 (https://gitweb.torproject.org/user/gk/tor-browser-bundle.git/commit/?h=bug_20301&id=dfa3403825fc315168df3295f6062f5979b0a8a0) in my public repo.

I applied the same fix, using the same comment, to tor-browser-build in branch bug_20301: https://gitweb.torproject.org/user/boklm/tor-browser-build.git/commit/?h=bug_20301&id=4e57f8a0780fd9ed9633dc7d2391bf908b74e337

The comment looks good to me. Although, we won't update GCC to 6.2.0 anymore. Thus, the commit message is wrong. I think we should just close this bug an apply the comment about disabling PIE in the upcoming patch for #25304 (moved).

Ok, I think it makes sense to do it in the patch for #25304 (moved).

Trac:
Resolution: N/A to duplicate
Status: needs_review to closed

closed

mentioned in issue #25304 (moved)

moved to tpo/applications/tor-browser#20301 (closed)