libera #seamonkey

10 Jun 2025

Surfing the net has never been so suite || Release: 2.53.21 | Beta: 2.53.21b1 seamonkey-project.org/releases || This channel is logged at logbot.thereisonlyxul.org/seamonkey || SeaMonkey Nightly 2.53 b1pre archive archive.seamonkey-project.org/nightly || c-c (OPEN ) , c-253 (APPROVAL ONLY) || R.I.P. WG9s

• 00:00
  frg_Away
  Compiling under x86 is totally unsupported by mozilla.
• 00:00
  Sompi
  first I need to figure out which package is the i386 cross-compilation C++ standard libraries
• 00:00
  nsITobin
  that is a thing go
• 00:00
  nsITobin
  the rust issue
• 00:01
  Sompi
  because that's what it is missing
• 00:01
  nsITobin
  and frg isn't wrong
• 00:02
  nsITobin
  but 32bit distros DO do it like that and always have..
• 00:02
  nsITobin
  because what mozilla supports used to not matter cause source code was easy to modify and build
• 00:03
  nsITobin
  you know Sompi that whole.. open source ecosystem shit no one likes anymore
• 00:03
  nsITobin
  :P
• 00:05
  Sompi
  it just complains that libstdc++ is not new enough
• 00:05
  Sompi
  probably means that it does not exist for that target
• 00:09
  Sompi
  it was lib64stdc++-12-dev-i386-cross
• 00:09
  Sompi
  now it is asking for the rust standard library
• 00:09
  nsITobin
  32bit rust
• 00:10
  Sompi
  how do I add them?
• 00:10
  Sompi
  the build system tells me to use "rustup" but I don't have that installed
• 00:11
  Sompi
  and I cannot find libstd-rust-cross-compilation packages from apt
• 00:11
  nsITobin
  you can use rustup but uninstall system rust
• 00:11
  nsITobin
  rustup.rs
• 00:12
  nsITobin
  gitlab.com/binaryoutcast/laboratori…up-init/package.spec?ref_type=heads
• 00:12
  Sompi
  is it telling me to pipe http payload to sh?
• 00:13
  nsITobin
  they removed the archive
• 00:13
  Sompi
  whatever
• 00:13
  Sompi
  it's a virtual machine
• 00:13
  Sompi
  who cares if it gets some weird shit in it
• 00:14
  Sompi
  the rust people really love to be in control of what the users of their programming language have in their computers
• 00:15
  nsITobin
  hey are fuckin cult members
• 00:15
  nsITobin
  cenobites
• 00:16
  Sompi
  is there also i386-unknown-linux-gnu?
• 00:16
  Sompi
  i586 seems to exist
• 00:17
  nsITobin
  does debian based still default to i386 not i586?
• 00:19
  Sompi
  I ran that rustup thing and now rust and cargo are somehow only runnable for root
• 00:19
  Sompi
  hmm
• 00:20
  Sompi
  I see, so it only installs it for the current user
• 00:20
  Sompi
  :|
• 00:20
  Sompi
  then how do I remove that
• 00:22
  Sompi
  now I have two installations of it, and I don't need it for root
• 00:23
  Sompi
  now it is compiling for i686-target
• 00:24
  Sompi
  it probably fails at linking but at least it is going somewhere
• 00:36
  Sompi
  sininenankka.dy.fi/kuvauppi/upload/q4rvq.jpg <- what is this error 2?
• 00:39
  nsITobin
  go up to the first error
• 00:39
  nsITobin
  that is only the final one
• 00:39
  Sompi
  but there's only warnings
• 00:40
  nsITobin
  error 2 as in make error 2
• 00:40
  nsITobin
  wait
• 00:40
  nsITobin
  you sure?
• 00:40
  nsITobin
  unless is this j1?
• 00:40
  nsITobin
  or only one core
• 00:41
  Sompi
  Doesn't it stop the build immediately when it encounters an error?
• 00:48
  Sompi
  apparently not
• 00:48
  Sompi
  I switched to single-core and now I see the error
• 00:49
  Sompi
  sininenankka.dy.fi/kuvauppi/upload/1e3zkp.jpg
• 00:49
  Sompi
  now I'm confused
• 00:51
  Sompi
  apparently the default size of a double is different for 32- and 64-bit platforms
• 00:52
  Sompi
  so I probably need some compiler flag
• 01:02
  Sompi
  nsITobin: I think the compiler options I need are -mdouble=64 and -mlong-double-128
• 01:02
  Sompi
  but the builder refuses to start the build because "Option must start with two dashes instead of one"
• 01:03
  nsITobin
  hmm
• 01:03
  Sompi
  But clang wants those with one -
• 01:03
  nsITobin
  ffs
• 01:13
  Sompi
  how do I solve that
• 01:13
  Sompi
  :(
• 02:09
  Sompi
  the python script seems to be written in such way that it is completely incompatible with compiler flags that begin with one -
• 02:10
  Sompi
  "sininenankka.dy.fi/kuvauppi/upload/n91yw.jpg
• 02:10
  Sompi
  sininenankka.dy.fi/kuvauppi/upload/n91yw.jpg
• 02:11
  Sompi
  that's a ridiculously bad way to do things
• 02:11
  Sompi
  sininenankka.dy.fi/kuvauppi/upload/136sml.jpg
• 02:12
  Sompi
  that is so bad that those functions there should be rewritten entirely
• 02:12
  Sompi
  it is also incompatible with compiler flargs that are not 100% lowercase
• 02:12
  Sompi
  seriously
• 02:13
  Sompi
  what the hell
• 02:14
  Sompi
  And if I patch it so that it accepts the flags that start with only one '-', some other function that joins the options afterwards loses the value of the compiler flag
• 02:38
  Sompi
  actually I think that the compiler flag I need is -ffp-eval-method=0
• 02:38
  Sompi
  but it also has only one - so it doesn't work
• 02:39
  Sompi
  They changed the floating point type definitions in some version of gcc headers and now it conflicts with the code in seamonkey's source
• 02:39
  nsITobin
  hmm.. optimize flags in mozconfig maybe?
• 02:42
  Sompi
  and that flag does not even exist in clang version 14.0.6 that Debian has... but it exists in version 20
• 02:42
  Sompi
  this is hopeless
• 02:44
  Sompi
  what if I just patch the header files directly...
• 02:44
  Sompi
  fuck gcc
• 02:44
  Sompi
  why it has to have different floating point type definitions for different x86 bitnesses
• 02:50
  Sompi
  they really like to break stuff
• 02:52
  Sompi
  and also redefining a standard data type like double_t in your own code is a very stupid thing to do, it is surely going to not work on every target platform where that data type is not already what you want it to be
• 02:53
  Sompi
  and somehow someone did that in mozilla's source code
• 02:54
  Sompi
  but anyway, it continues compiling now after I changed the definition of __GLIBC_FLT_EVAL_METHOD for other than 64-bit x86 targets to 0 in /usr/i686-linux-gnu/include/bits/flt-eval-method.h
• 02:54
  Sompi
  So the entire SeaMonkey codebase depends on it being 0
• 02:55
  Sompi
  and by default it is 2 for 32-bit targets, and only newer versions of clang allow changing it from the command line
• 08:45
  njsg
  Tobin: swapping wouldn't be an issue here (other than possible thrashing, but this being a VM maybe it was all cached in RAM?), the 4 GiB total is *with* swap.
• 08:49
  njsg
  Sompi: did you start a clean build when you changed to cross-compiling?
• 12:17
  frg
  Linux and macOS cross build scripts and configs from the Linux builder:
• 12:17
  frg
  seafiles.thereisonlyxul.org/people/frg/build.zip
• 12:17
  frg
  Obviously the google api keys are not included so you need to provide your own or remove them from the configs. Only used for smartscreen and location services.
• 12:18
  frg
  Also obviously they won't work without adapting the path statements too.
• 12:18
  frg
  gcc 8.50 /clang 8 (llvm) and clang 14 .06 cross for macOS only.
• 12:41
  frg
  And here the Windows ones: seafiles.thereisonlyxul.org/people/frg/mozconfigs-25321-win.zip
• 13:49
  Sompi
  njsg: How do I make sure that it starts a clean build?
• 14:10
  frg_Away
  mach clobber or delete the object directory
• 14:20
  nsITobin
  Good morning
• 14:20
  frg_Away
  hi nsITobin
• 15:36
  tomman
  the new hotness on DNS: HTTPS records
• 15:36
  tomman
  rohanrd.xyz/posts/hosting-website-on-phone
• 15:36
  tomman
  kalfeher.com/https-current-state
• 15:37
  tomman
  i guess we at Mozillaland don't have to worry that much yet: only Safari seems to properly support those records
• 15:37
  tomman
  Chromovision support is half-assed, and Firefox support is even more dire
• 15:37
  tomman
  and only cloud provider DNS services offer easy interfaces to manage said entries
• 15:38
  tomman
  on the flip side, it doesn't seem to rely on D'OH!, as it can work over traditional DNS too