Opened 2 months ago

Closed 4 weeks ago

Last modified 4 weeks ago

#31449 closed defect (worksforme)

Signing tools for 32bit Linux are 64bit now

Reported by: gk Owned by: tbb-team
Priority: High Milestone:
Component: Applications/Tor Browser Version:
Severity: Normal Keywords: tbb-rbm, tbb-9.0-must-alpha, TorBrowserTeam201909
Cc: boklm Actual Points: 1
Parent ID: #30321 Points: 0.25
Reviewer: Sponsor:

Description

It seems our signing tools we build when compiling 32bit Tor Browser are 64bit now, but we need 32bit ones.

Child Tickets

Change History (14)

comment:1 Changed 2 months ago by gk

Parent ID: #30321

comment:2 Changed 8 weeks ago by gk

Keywords: tbb-9.0-must-alpha added; tbb-9.0-alpha-must removed

To be more precise here: only mar and msbdiff are 64-bit now while the other binary files we bundle are 32-bit as expected.

Last edited 8 weeks ago by gk (previous) (diff)

comment:3 Changed 7 weeks ago by boklm

Status: newneeds_information

After looking at this issue, it seems that:

However it seems that the only place where mbsdiff is used is for generating incremental mar files, which we usually do on a 64bit machine, so I'm not sure we need a 32bit mbsdiff. It seems also that our signing script is only using signmar and not mar. Is there an other reason that I'm forgetting for wanting 32bit versions of mbsdiff and mar?

comment:4 in reply to:  3 ; Changed 7 weeks ago by gk

Status: needs_informationnew

Replying to boklm:

After looking at this issue, it seems that:

  • signmar is the same as mar with some additional features. So if we need a 32 bit mar, we could probably use signmar.

If we could verify that and/or be sure that signmar is not calling mar under the hood that would be good.

However it seems that the only place where mbsdiff is used is for generating incremental mar files, which we usually do on a 64bit machine, so I'm not sure we need a 32bit mbsdiff. It seems also that our signing script is only using signmar and not mar. Is there an other reason that I'm forgetting for wanting 32bit versions of mbsdiff and mar?

Fair enough. I was under the impression that the current state risks our mar signing AND that we do something wrong with our 32bit builds given that the issue did not happen in the past and we already seem to need some hacks to get 32bit builds going. If neither is the case, great! We could closed this ticket then as WORKSFORME or something similar.

comment:7 in reply to:  4 ; Changed 7 weeks ago by boklm

Replying to gk:

Fair enough. I was under the impression that the current state risks our mar signing AND that we do something wrong with our 32bit builds given that the issue did not happen in the past and we already seem to need some hacks to get 32bit builds going. If neither is the case, great! We could closed this ticket then as WORKSFORME or something similar.

I think that we did not have this issue in the past because our .mozconfig-linux-i686 file included the line ac_add_options --host=i686-linux-gnu which caused host tools to be built in 32bit. Although that was not really correct as the host was in reality 64bit, but it still worked as we can run 32bit binaries on a 64bit host.

comment:8 in reply to:  7 Changed 7 weeks ago by gk

Replying to boklm:

Replying to gk:

Fair enough. I was under the impression that the current state risks our mar signing AND that we do something wrong with our 32bit builds given that the issue did not happen in the past and we already seem to need some hacks to get 32bit builds going. If neither is the case, great! We could closed this ticket then as WORKSFORME or something similar.

I think that we did not have this issue in the past because our .mozconfig-linux-i686 file included the line ac_add_options --host=i686-linux-gnu which caused host tools to be built in 32bit. Although that was not really correct as the host was in reality 64bit, but it still worked as we can run 32bit binaries on a 64bit host.

Makes sense. What is Mozilla doing for their 32bit builds? Do they get the same result as we? If so, it might be worth filing an upstream bug. If not, we should fix our setup.

comment:9 Changed 6 weeks ago by gk

As another data point: I looked at the mar-tools for Windows and macOS we currently get for 9.0a4. And there both mar and msbdiff are 64bit ELF binaries. So, that seems to be "normal" for cross-compiling...

comment:10 Changed 6 weeks ago by cypherpunks

As another data point: I looked at the mar-tools for Windows and macOS we currently get for 9.0a4. And there both mar and msbdiff are 64bit ELF binaries. So, that seems to be "normal" for cross-compiling...

If they should be executed on host only (and not on target), then sure!

comment:11 Changed 6 weeks ago by gk

Keywords: TorBrowserTeam201909 added; TorBrowserTeam201908 removed

Moving must-alpha tickets to September.

comment:12 Changed 6 weeks ago by pili

Points: 0.25

comment:13 Changed 4 weeks ago by gk

Resolution: worksforme
Status: newclosed

9.0a6 got signed with the mar-tools we produced, so things are working. I think that means wfm.

comment:14 Changed 4 weeks ago by boklm

Actual Points: 1
Note: See TracTickets for help on using tickets.