closure: add ppc64 ELFv1 support

[tankf33der]

Second step to add ppc64 support for closure.
Patch manually tested on x64 and ppc64.

Only ELFv2 has been supported all this time — now and always — and the assembly code for each architecture operated under this memory model. Because of ppc64, additional code had to be written for the closure itself to make it work with ELFv1 as well. All closure tests pass, which is a very good sign to start with.

tankf33der@cfarm121:~/v$ VTEST_ONLY=closure v test vlib/
---- Testing... ----------------------------------------------------------------------------------------------------------------------------
 OK    [ 1/31] C:  1101.3 ms, R:     4.721 ms vlib/v/tests/fns/closure_in_if_guard_2_test.v
 OK    [ 2/31] C:  1113.0 ms, R:     5.135 ms vlib/v/tests/comptime/comptime_call_method_closure_test.v
 OK    [ 3/31] C:  1119.0 ms, R:     5.209 ms vlib/v/tests/generics/generics_return_closure_test.v
 OK    [ 4/31] C:  1132.7 ms, R:     4.056 ms vlib/v/tests/closure_with_fn_ref_var_test.v
 OK    [ 5/31] C:  1156.2 ms, R:     5.232 ms vlib/v/tests/builtin_arrays/array_filter_using_direct_closure_test.v
 OK    [ 6/31] C:  1152.0 ms, R:     3.351 ms vlib/v/tests/generics/return_closure_f64_test.v
 OK    [ 7/31] C:  1218.4 ms, R:     4.042 ms vlib/v/tests/arrays_closure_fixed_array_test.v
 OK    [ 8/31] C:  1215.0 ms, R:     3.345 ms vlib/v/tests/generics/generics_closure_fn_direct_call_test.v
 OK    [ 9/31] C:  1239.8 ms, R:     5.390 ms vlib/v/tests/fns/closure_in_if_guard_1_test.v
 OK    [10/31] C:  1248.1 ms, R:     3.859 ms vlib/v/tests/assign/assign_literal_with_closure_test.v
 OK    [11/31] C:  1245.8 ms, R:     3.190 ms vlib/v/tests/generics/generics_closures_with_different_generic_types_test.v
 OK    [12/31] C:  1254.8 ms, R:     7.260 ms vlib/v/tests/fns/go_call_anon_fn_with_closure_test.v
 OK    [13/31] C:  1294.2 ms, R:     4.208 ms vlib/v/tests/generics/generics_closure_fn_test.v
 OK    [14/31] C:  1301.2 ms, R:     4.451 ms vlib/v/tests/fns/closure_fn_arg_in_map_test.v
 OK    [15/31] C:  1299.2 ms, R:     3.548 ms vlib/v/tests/interfaces/interface_closure_test.v
 OK    [16/31] C:  1302.1 ms, R:     2.675 ms vlib/v/tests/match_multi_return_closure_heap_test.v
 OK    [17/31] C:  1320.7 ms, R:     3.680 ms vlib/v/tests/comptime/comptime_closure_field_access_test.v
 OK    [18/31] C:  1362.9 ms, R:     3.934 ms vlib/v/tests/fns/closure_with_fixed_array_var_test.v
 OK    [19/31] C:  1367.2 ms, R:     4.048 ms vlib/v/tests/fns/closure_data_with_gc_test.c.v
 OK    [20/31] C:  1367.4 ms, R:     2.712 ms vlib/v/tests/fns/closure_option_direct_call_test.v
 OK    [21/31] C:  1408.2 ms, R:     3.160 ms vlib/v/tests/fns/anon_c_keywords_closure_test.v
 OK    [22/31] C:  1416.1 ms, R:     3.153 ms vlib/v/tests/fns/closure_variable_in_smartcast_test.v
 OK    [23/31] C:  1420.2 ms, R:     2.884 ms vlib/v/tests/fns/closure_of_method_defined_on_alias_test.v
 OK    [24/31] C:  1463.0 ms, R:     3.556 ms vlib/v/tests/generics/return_closure_int_test.v
 OK    [25/31] C:  1465.0 ms, R:     2.590 ms vlib/v/tests/interfaces/interface_method_closure_test.v
 OK    [26/31] C:  1472.9 ms, R:     3.664 ms vlib/v/tests/fns/closure_of_instance_method_passed_to_voidptr_parameter_test.v
 OK    [27/31] C:  1487.2 ms, R:     2.840 ms vlib/v/tests/builtin_arrays/array_ops_create_just_one_closure_test.c.v
 OK    [28/31] C:  1508.1 ms, R:     3.894 ms vlib/v/tests/fns/closure_with_sumtype_var_test.v
 OK    [29/31] C:  1518.8 ms, R:     2.775 ms vlib/v/tests/fns/closure_test.v
 OK    [30/31] C:  1691.8 ms, R:     4.607 ms vlib/v/tests/fns/closure_struct_init_cov_regression_test.v
 OK    [31/31] C:  1309.4 ms, R:  3490.082 ms vlib/v/tests/fns/closure_generator_test.v
--------------------------------------------------------------------------------------------------------------------------------------------
Summary for all V _test.v files: 31 passed, 31 total. Elapsed time: 4829 ms, on 31 parallel jobs. Comptime: 40971 ms. Runtime: 3607 ms.
tankf33der@cfarm121:~/v$ uname  -a
Linux cfarm121 6.19.6+deb14-powerpc64 #1 SMP PREEMPT Debian 6.19.6-1 (2026-03-06) ppc64 GNU/Linux
tankf33der@cfarm121:~/v$ file ./v
./v: ELF 64-bit MSB pie executable, 64-bit PowerPC or cisco 7500, Power ELF V1 ABI, version 1 (SYSV), dynamically linked, interpreter /lib64/ld64.so.1, BuildID[sha1]=381f6b9ad31819ac0d4b694fbb1826ea55e1324d, for GNU/Linux 3.2.0, not stripped

Note Nr1337 for future me — ppc64 has this peculiarity that in our galaxy there exist versions for two ELF versions. If someone ever requests ppc64 support in ELFv2, add comptime support for elfv1 ($if elfv1 ...) and just reverse the stub from ppc64le. Get access to a system for testing to be sure idea is ok.

[tankf33der]

@JalonSolov FYI, CI must be green.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a35793e6ce

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Keep first closure thunk variant longest

The file-level note says translated/bootstrap C mode (vc/v.c) infers a fixed C array size from the first $if branch, but this new ppc64 thunk is longer than the first ppc64le thunk. On ELFv1 ppc64 bootstrap builds, that mismatch can produce an undersized initializer (or truncated thunk bytes), which breaks closure trampolines at runtime. Please make the first branch the longest one again (or pad it) so bootstrap sizing remains safe.

Useful? React with 👍 / 👎.

[JalonSolov]

Some issues with native backend.

[JalonSolov]

Bounce for final check.