LibEvent Build for Android
Activity
For Reference, Google's Android build config
I verified it builds on 4 targets. I updated the version to 2.1.11
https://github.com/sisbell/tor-browser-build/commit/da222654e20ad1878f238fde0eb1f9caed905d53
If we continue with 2.1.8, we get the following error. We can apply a patch if we want to support this older version.
./.libs/libevent_core-2.1.so: error: undefined reference to 'arc4random_addrandom'I have the other flags because sample and libevent regress are triggering tests which have unsupported classes (older versions of tor-android do the same thing)
test/test-dumpevents.c:104: error: undefined reference to 'stdout' test/test-dumpevents.c:104: error: undefined reference to 'stderr'If we need libevent regress, we could patch and remove the tests
AM_CONDITIONAL([BUILD_SAMPLES], [test "$enable_samples" = "yes"]) AM_CONDITIONAL([BUILD_REGRESS], [test "$enable_libevent_regress" = "yes"])Trac:
Status: assigned to needs_review
Keywords: TorBrowserTeam201912 deleted, TorBrowserTeam201912R added
FWIW, I think bumping Libevent to 2.1.11 is the right thing to do (we would squash #31499 (moved) while doing so, which is nice).
Replying to [ticket:28704#comment:23 sisbell]:
Other issues:
- There was a suggestion to move some of the fields in configure_opt up to rbm. OpenSSL doesn't use the same configure_host value as other projects so this will require some more discussion if we want to move forward with this suggestion.
- Information regarding libevent --disable-libevent-regress--disable-samples. I need to look back through my notes. I'll post in a follow up comment.
From ticket:28704#comment:26, do including these change the size of the resulting library? I'd rather disable these for all platforms in a separate ticket.
Replying to sysrqb:
Replying to [ticket:28704#comment:23 sisbell]:
Other issues:
- There was a suggestion to move some of the fields in configure_opt up to rbm. OpenSSL doesn't use the same configure_host value as other projects so this will require some more discussion if we want to move forward with this suggestion.
- Information regarding libevent --disable-libevent-regress--disable-samples. I need to look back through my notes. I'll post in a follow up comment.
From ticket:28704#comment:26, do including these change the size of the resulting library? I'd rather disable these for all platforms in a separate ticket.
The tar files themselves are a few bytes off (looks to be due to timestamp differences on files). But after unarchiving, the following directories all have the same size
- libevent --disable-libevent-regress --disable-samples
- libevent --disable-libevent-regress
- libevent --disable-samples
- libevent
Broke commit into its own branch, otherwise no changes since last commit.
https://github.com/sisbell/tor-browser-build/commits/bug-28765
Trac:
Status: needs_revision to needs_review
Replying to sisbell:
Broke commit into its own branch, otherwise no changes since last commit.
https://github.com/sisbell/tor-browser-build/commits/bug-28765
When looking at commit f73f0125252cdf18596291c45ba3d92cb719b883:
You define
var/configure_optto--disable-staticinproject/libevent/configfor all platforms except android. I think this is wrong asvar/configure_optalready has some values defined inrbm.conf, so we override those values which will probably break the build on non-android platforms.By the way in ticket:28704#comment:14 I said this:
it seems we could have a var/configure_opt for android in
rbm.confcontaining something likeCC=clang --host=[% c("var/host") %] [% c("var/configure_opt_project") %], wherevar/configure_opt_projectis defined in each project to define options specific to this project I still think that would be a good idea. If you don't think so, then please at least explain why instead of just ignoring comments and not doing the changes suggested.Regarding update to version
2.1.11, I think it would be a good idea to mention it in the commit message. Or maybe do it as a separate commit (with #31499 (moved) as the bug number) as this is a change that affect all platforms.Trac:
Status: needs_review to needs_revision
Keywords: TorBrowserTeam202002R deleted, TorBrowserTeam202002 addedReplying to boklm:
Replying to sisbell:
Broke commit into its own branch, otherwise no changes since last commit.
https://github.com/sisbell/tor-browser-build/commits/bug-28765
When looking at commit f73f0125252cdf18596291c45ba3d92cb719b883:
You define
var/configure_optto--disable-staticinproject/libevent/configfor all platforms except android. I think this is wrong asvar/configure_optalready has some values defined inrbm.conf, so we override those values which will probably break the build on non-android platforms. Ok that should be easy to fix.By the way in ticket:28704#comment:14 I said this:
it seems we could have a var/configure_opt for android in
rbm.confcontaining something likeCC=clang --host=[% c("var/host") %] [% c("var/configure_opt_project") %], wherevar/configure_opt_projectis defined in each project to define options specific to this projectI still think that would be a good idea. If you don't think so, then please at least explain why instead of just ignoring comments and not doing the changes suggested.
I didn't ignore the comment. i previously stated the reason I didn't do it and requested more discussion around a solution.
"There was a suggestion to move some of the fields in configure_opt up to rbm. OpenSSL doesn't use the same configure_host value as other projects so this will require some more discussion if we want to move forward with this suggestion."
https://trac.torproject.org/projects/tor/ticket/28704#comment:23
Regarding update to version
2.1.11, I think it would be a good idea to mention it in the commit message. Or maybe do it as a separate commit (with #31499 (moved) as the bug number) as this is a change that affect all platforms.-
Replying to sisbell:
By the way in ticket:28704#comment:14 I said this:
it seems we could have a var/configure_opt for android in
rbm.confcontaining something likeCC=clang --host=[% c("var/host") %] [% c("var/configure_opt_project") %], wherevar/configure_opt_projectis defined in each project to define options specific to this projectI still think that would be a good idea. If you don't think so, then please at least explain why instead of just ignoring comments and not doing the changes suggested.
I didn't ignore the comment. i previously stated the reason I didn't do it and requested more discussion around a solution.
Ah, indeed, I missed that. Sorry.
"There was a suggestion to move some of the fields in configure_opt up to rbm. OpenSSL doesn't use the same configure_host value as other projects so this will require some more discussion if we want to move forward with this suggestion."
I think openssl not using
var/configure_optis not a reason for not using it in other projects. We are in the same situation on other platforms, where we definevar/configure_opt, and use it in some projects, and not in others projects expecting different options like openssl. -
Replying to boklm:
Replying to sisbell:
Broke commit into its own branch, otherwise no changes since last commit.
https://github.com/sisbell/tor-browser-build/commits/bug-28765
When looking at commit f73f0125252cdf18596291c45ba3d92cb719b883:
You define
var/configure_optto--disable-staticinproject/libevent/configfor all platforms except android. I think this is wrong asvar/configure_optalready has some values defined inrbm.conf, so we override those values which will probably break the build on non-android platforms.By the way in ticket:28704#comment:14 I said this:
it seems we could have a var/configure_opt for android in
rbm.confcontaining something likeCC=clang --host=[% c("var/host") %] [% c("var/configure_opt_project") %], wherevar/configure_opt_projectis defined in each project to define options specific to this project I still think that would be a good idea. If you don't think so, then please at least explain why instead of just ignoring comments and not doing the changes suggested.Regarding update to version
2.1.11, I think it would be a good idea to mention it in the commit message. Or maybe do it as a separate commit (with #31499 (moved) as the bug number) as this is a change that affect all platforms.+1 for a separate commit before the patch for this ticket (and making sure the other platforms are not exploding)
#31499 (moved) now includes the version change for libevent (I sitll need to test this for non-android platforms)
#33216 (moved) contains the CC=clang and host info
This commit now includes using a configure_opt_project variable. I verified this is building correctly for android. Again I still need to test with #31499 (moved) to verify all platforms.
!https://github.com/sisbell/tor-browser-build/commits/bug-28765a
-
Replying to sisbell:
This commit now includes using a configure_opt_project variable. I verified this is building correctly for android. Again I still need to test with #31499 (moved) to verify all platforms.
I think it doesn't work on other platforms: the commit moves the
--disable-staticfromprojects/libevent/buildtovar/configure_opt_project, but thevar/configure_optfrom other platforms does not usevar/configure_opt_project.I think we should update the
var/configure_optinrbm.conffrom all platforms to includevar/configure_opt_project.Trac:
Keywords: TorBrowserTeam202003R deleted, TorBrowserTeam202003 added
Status: needs_review to needs_revision I updated the other platforms to includevar/configure_opt_projectin the rbm.conf file. Built windows, mac and linux variant to make sure it didn't break
https://github.com/sisbell/tor-browser-build/commits/bug-28765b
Trac:
Status: needs_revision to needs_review
Keywords: TorBrowserTeam202003 deleted, TorBrowserTeam202003R added-
Replying to sysrqb:
Is
configure_opt_projectincluded on linux-x86_64?Yes, I think it is missing.
By the way, while doing this change, I think we can simplify
configure_optonlinux-i686by renamingvar/configure_opt_i686tovar/configure_opt. I'm not sure why we have thisconfigure_opt_i686, as it doesn't seem to be used anywhere else. I added the 'configure_opt_project' in rbm.conf for linux-x86_64 so that it will pick now
I removed reference to 'var/configure_opt_i686' and just collapsed it into the 'var/configure_opt' field
https://github.com/sisbell/tor-browser-build/commits/bug-28765c
Trac:
Status: needs_revision to needs_review
Keywords: TorBrowserTeam202004 deleted, TorBrowserTeam202004R added-
Replying to sisbell:
I added the 'configure_opt_project' in rbm.conf for linux-x86_64 so that it will pick now
I removed reference to 'var/configure_opt_i686' and just collapsed it into the 'var/configure_opt' field
https://github.com/sisbell/tor-browser-build/commits/bug-28765c
The patch for #28765 (moved) looks good, but is difficult to test since it is based on a version of the patch for #31499 (moved) which doesn't build on Linux.
Trac:
Keywords: TorBrowserTeam202004R deleted, TorBrowserTeam202004 added
Status: needs_review to needs_revision I made the changes to #31499 (moved) so this is ready for review again.
Trac:
Status: needs_revision to needs_review
Keywords: TorBrowserTeam202004 deleted, TorBrowserTeam202004R addedAfter breaking out #33877 (moved), this commit only contains the android config
https://github.com/sisbell/tor-browser-build/commits/bug-28765d
mentioned in issue #31499 (moved)
mentioned in issue #33877 (moved)
