Re: [zstd-sys 2.0.1+zstd.1.5.2] crate failing on arm64
Date: Thu, 24 Nov 2022 17:26:51 UTC
On Nov 24, 2022, at 07:00, Mark Millard <marklmi@yahoo.com> wrote:
> On Nov 24, 2022, at 03:03, Nuno Teixeira <eduardo@freebsd.org> wrote:
>
>> Hello Mark,
Hello.
>> I have compared some of errors/warnings with amd64 build logs and they are present in there too.
Interesting.
>> I think I found a glitch at the end of arm64 log:
>> ---
>> [zstd-sys 2.0.1+zstd.1.5.2] running: "ar" "cq" "/wrkdirs/usr/ports...
>> (...)
>> "/wrkdirs/usr/ports/editors/lapce/
>> work/target/aarch64-unknown-freebsd/release/build/zstd-sys-97d70ebd740964f8/out/zstd/lib/decompress/huf_decompress_amd64.o"
>> ^^^^^
Well, in my older context,
# grep "^#" /wrkdirs/usr/ports/editors/lapce/work/lapce-0.2.1/cargo-crates/zstd-sys-2.0.1+zstd.1.5.2/zstd/lib/decompress/huf_decompress_amd64.S | grep -v "#define "
#include "../common/portability_macros.h"
#if defined(__ELF__) && defined(__GNUC__)
#endif
#if ZSTD_ENABLE_ASM_X86_64_BMI2
#undef GET_NEXT_DELT
#undef DECODE_FROM_DELT
#undef DECODE
#undef RELOAD_BITS
#undef DECODE
#undef RELOAD_BITS
#endif
The #endif is the last line of the file and, ignoring
comments and blank lines for the most part, the rest
of the file before/at the #if ZSTD_ENABLE_ASM_X86_64_BMI2
is just:
#include "../common/portability_macros.h"
/* Stack marking
* ref: https://wiki.gentoo.org/wiki/Hardened/GNU_stack_quickstart
*/
#if defined(__ELF__) && defined(__GNUC__)
.section .note.GNU-stack,"",%progbits
#endif
#if ZSTD_ENABLE_ASM_X86_64_BMI2
Looks like the ZSTD_ENABLE_ASM_X86_64_BMI2 test would eliminate the
amd64 specifics in an aarch64 context, so they do not bother avoiding
the file.
>> ---
>> and zstd-sys-2.0.1+zstd.1.5.2/zstd/lib/common/xxhash.h:
>> # if (defined(__aarch64__) || defined(__arm64__) || defined(_M_ARM64) || defined(_M_ARM64EC)) \
>>
>> So I presume that this crate should be build on arm64/aarch64 but don't understant why it calls:
>> "huf_decompress_amd64.o"
>>
>> Any clues?
>
> Not at this point. I've got the system rebuilding the
> port so I can set up to look again.
>
>> . . .
FYI: the better string for searching got:
# find -s /wrkdirs/usr/ports/editors/lapce/ -name stderr -exec grep -l "aborting due to" {} \; | less
/wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/cap-fs-ext-10fe2c1f66453130/stderr
/wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/cap-primitives-ed08064314a4640b/stderr
/wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/cap-std-5acaec63374cb836/stderr
/wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/io-extras-e83e1591d250cc25/stderr
/wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/io-lifetimes-62b7366622512d7e/stderr
/wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/system-interface-56dbb6efd7f0321e/stderr
The first line of the output is new compared to before:
# more /wrkdirs/usr/ports/editors/lapce/work/target/aarch64-unknown-freebsd/debug/build/cap-fs-ext-10fe2c1f66453130/stderr
error[E0554]: `#![feature]` may not be used on the stable release channel
--> <anon>:1:12
|
1 | #![feature(windows_by_handle)]
| ^^^^^^^^^^^^^^^^^
error[E0635]: unknown feature `windows_by_handle`
--> <anon>:1:12
|
1 | #![feature(windows_by_handle)]
| ^^^^^^^^^^^^^^^^^
error: aborting due to 2 previous errors
Some errors have detailed explanations: E0554, E0635.
For more information about an error, try `rustc --explain E0554`.
So, not a new type of error compared to the others.
===
Mark Millard
marklmi at yahoo.com