Big increases in file size after v0.2.16

[curiousdannii]

I was updating the version of Jiff I use, and noticed that since v0.2.16 there have been some really big increases in file sizes. These are the final wasm sizes for my app (release build). No changes to my code other than changing which version of Jiff I use.

Jiff Version	Final WASM size
v0.2.15	624KB
v0.2.16	625KB
v0.2.17-0.2.18	716KB
v0.2.19	731KB
v0.2.20-0.2.24	740KB

That's an increase of 115KB or 19% of my final binary size purely from updating Jiff. Ironically the changelog for v0.2.17 says it "contains binary size improvements to Jiff"!

Rust version = 1.94.0
Target = wasm32-unknown-emscripten

I'm using Jiff with default-features = false.

[curiousdannii]

Twiggy (complete) results for v0.2.16:

 Shallow Bytes │ Shallow % │ Item
          4456 ┊     0.71% ┊ <jiff::civil::datetime::DateTimeArithmetic>::checked_add (.1041)
          3699 ┊     0.59% ┊ <jiff::civil::date::DateArithmetic>::checked_add (.1042)
          3696 ┊     0.59% ┊ <jiff::civil::date::DateArithmetic>::checked_add
          1545 ┊     0.25% ┊ <jiff::shared::PosixDayTime>::to_datetime
          1241 ┊     0.20% ┊ <jiff::shared::posix::DstInfo<&str>>::in_dst
           720 ┊     0.12% ┊ <jiff::shared::util::itime::IDate>::checked_add_days
           548 ┊     0.09% ┊ <jiff::civil::time::Time>::overflowing_add_duration_general
           305 ┊     0.05% ┊ <jiff::shared::util::itime::IDateTime>::saturating_add_seconds
           269 ┊     0.04% ┊ <jiff::error::Error as core::fmt::Debug>::fmt
           201 ┊     0.03% ┊ <jiff::error::ErrorKind as core::fmt::Debug>::fmt
           182 ┊     0.03% ┊ <jiff::error::ErrorKind as core::fmt::Display>::fmt
            54 ┊     0.01% ┊ <&jiff::shared::util::error::Error as core::fmt::Debug>::fmt
            54 ┊     0.01% ┊ <&jiff::error::RangeError as core::fmt::Debug>::fmt
            51 ┊     0.01% ┊ <jiff::shared::util::error::Error as core::fmt::Debug>::fmt
            46 ┊     0.01% ┊ <jiff::error::Error>::adhoc_from_args
            31 ┊     0.00% ┊ <jiff::civil::datetime::DateTime>::checked_add_span::{closure#0}
            26 ┊     0.00% ┊ <&jiff::error::FilePathError as core::fmt::Debug>::fmt
            24 ┊     0.00% ┊ <&jiff::error::AdhocError as core::fmt::Debug>::fmt

Twiggy results for v0.2.24:

 Shallow Bytes │ Shallow % │ Item
         22317 ┊     3.02% ┊ <jiff::error::Error as core::fmt::Debug>::fmt
          6770 ┊     0.92% ┊ <jiff::civil::datetime::DateTimeArithmetic>::checked_add (.1053)
          3612 ┊     0.49% ┊ <jiff::civil::date::DateArithmetic>::checked_add (.1054)
          3603 ┊     0.49% ┊ <jiff::civil::date::DateArithmetic>::checked_add
          1993 ┊     0.27% ┊ <&jiff::error::fmt::temporal::Error as core::fmt::Debug>::fmt
          1710 ┊     0.23% ┊ <&jiff::util::b::BoundsError as core::fmt::Debug>::fmt
          1550 ┊     0.21% ┊ <jiff::shared::PosixDayTime>::to_datetime
          1382 ┊     0.19% ┊ <jiff::util::escape::Bytes as core::fmt::Display>::fmt
          1294 ┊     0.17% ┊ <jiff::error::ErrorKind as core::fmt::Debug>::fmt
          1252 ┊     0.17% ┊ <&jiff::error::fmt::rfc2822::Error as core::fmt::Debug>::fmt
          1241 ┊     0.17% ┊ <jiff::shared::posix::DstInfo<&str>>::in_dst
          1222 ┊     0.17% ┊ <&jiff::error::fmt::strtime::Error as core::fmt::Debug>::fmt
          1198 ┊     0.16% ┊ <&jiff::error::fmt::strtime::ParseError as core::fmt::Debug>::fmt
           799 ┊     0.11% ┊ <jiff::shared::posix::PosixDateError as core::fmt::Display>::fmt
           718 ┊     0.10% ┊ <&jiff::error::fmt::offset::Error as core::fmt::Debug>::fmt
           698 ┊     0.09% ┊ <jiff::shared::util::itime::IDate>::checked_add_days
           577 ┊     0.08% ┊ <&jiff::error::fmt::util::Error as core::fmt::Debug>::fmt
           520 ┊     0.07% ┊ <&jiff::error::unit::UnitConfigError as core::fmt::Debug>::fmt
           512 ┊     0.07% ┊ <&jiff::error::tz::offset::Error as core::fmt::Debug>::fmt
           498 ┊     0.07% ┊ <&jiff::error::fmt::rfc9557::Error as core::fmt::Debug>::fmt
           496 ┊     0.07% ┊ <jiff::civil::time::Time>::overflowing_add_duration_general
           479 ┊     0.06% ┊ <jiff::tz::offset::Offset as core::fmt::Display>::fmt
           427 ┊     0.06% ┊ <jiff::tz::timezone::repr::Repr as core::fmt::Debug>::fmt
           412 ┊     0.06% ┊ <&jiff::error::span::Error as core::fmt::Debug>::fmt
           411 ┊     0.06% ┊ <jiff::shared::util::itime::RangeError as core::fmt::Debug>::fmt

And then there's loads more, including 53 jiff::util::b::RawBoundsError functions, each 140-160 bytes.

So this is basically another repeat of #438, but it's gone massively backwards since then.

I think for now I'll pin v0.2.16.

[BurntSushi]

Can you provide a complete reproduction here? When I run twiggy on an x86-64 binary, I see completely different results:

$ twiggy top ./target/release/bttf -n 20
Warning: Couldn't find any relocations. The dominators, garbage and paths subcommands will not function correctly.
Hint: Try recompiling the binary with --emit-relocs.

 Shallow Bytes │ Shallow % │ Item
───────────────┼───────────┼─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
         70512 ┊     1.62% ┊ anon.6a54bf65387e1713d1f89389b13c18de.3023.llvm.11933177387520447906
         16111 ┊     0.37% ┊ <jiff[b78f59b61f1a8a0]::fmt::strtime::printer::Formatter<jiff[b78f59b61f1a8a0]::fmt::strtime::DefaultCustom>>::format::{closure#0}
         15680 ┊     0.36% ┊ <std[5264d755c47072bd]::backtrace_rs::symbolize::gimli::Cache>::with_global::<std[5264d755c47072bd]::backtrace_rs::symbolize::gimli::resolve::{closure#1}>
         15580 ┊     0.36% ┊ <jiff[b78f59b61f1a8a0]::fmt::strtime::parse::Parser>::parse
         13415 ┊     0.31% ┊ <bttf[9e54f3c1c322472b]::ical::RecurrenceIter as core[435d43ddfce3abfa]::iter::traits::iterator::Iterator>::next
         13316 ┊     0.31% ┊ regex_automata[9866b3dd34b7d80]::meta::strategy::new
         13015 ┊     0.30% ┊ <std[5264d755c47072bd]::backtrace_rs::symbolize::gimli::Context>::new
         11678 ┊     0.27% ┊ <regex_automata[9866b3dd34b7d80]::meta::strategy::ReverseInner as regex_automata[9866b3dd34b7d80]::meta::strategy::Strategy>::search_slots
         11524 ┊     0.26% ┊ <regex_syntax[dee4559d4c856994]::hir::translate::TranslatorI as regex_syntax[dee4559d4c856994]::ast::visitor::Visitor>::visit_post
         10685 ┊     0.25% ┊ regex_automata[9866b3dd34b7d80]::hybrid::search::find_fwd
         10549 ┊     0.24% ┊ <regex_automata[9866b3dd34b7d80]::nfa::thompson::compiler::Compiler>::c
          9920 ┊     0.23% ┊ zmij[941a6f4457105d6f]::POW10_SIGNIFICANDS
          9876 ┊     0.23% ┊ <regex_automata[9866b3dd34b7d80]::nfa::thompson::pikevm::PikeVM>::search_imp
          9789 ┊     0.22% ┊ <regex_automata[9866b3dd34b7d80]::meta::strategy::ReverseSuffix as regex_automata[9866b3dd34b7d80]::meta::strategy::Strategy>::search_slots
          9519 ┊     0.22% ┊ <aho_corasick[5b115f1abdb387d2]::dfa::Builder>::build_from_noncontiguous
          9517 ┊     0.22% ┊ <regex_automata[9866b3dd34b7d80]::nfa::thompson::pikevm::PikeVM>::which_overlapping_imp
          9085 ┊     0.21% ┊ <gimli[5191a9b065fcd249]::read::dwarf::Unit<gimli[5191a9b065fcd249]::read::endian_slice::EndianSlice<gimli[5191a9b065fcd249]::endianity::LittleEndian>, usize>>::new
          8401 ┊     0.19% ┊ <core[435d43ddfce3abfa]::cell::once::OnceCell<core[435d43ddfce3abfa]::result::Result<addr2line[dde38b7dfaeba16d]::line::Lines, gimli[5191a9b065fcd249]::read::Error>>>::try_init::<<core[435d43ddfce3abfa]::cell::once::OnceCell<core[435d43ddfce3abfa]::result::Result<addr2line[dde38b7dfaeba16d]::line::Lines, gimli[5191a9b065fcd249]::read::Error>>>::get_or_init<<addr2line[dde38b7dfaeba16d]::line::LazyLines>::borrow<gimli[5191a9b065fcd249]::read::endian_slice::EndianSlice<gimli[5191a9b065fcd249]::endianity::LittleEndian>>::{closure#0}>::{closure#0}, !>
          8072 ┊     0.19% ┊ <regex_syntax[dee4559d4c856994]::hir::translate::TranslatorI as regex_syntax[dee4559d4c856994]::ast::visitor::Visitor>::visit_class_set_item_post
          7671 ┊     0.18% ┊ <jiff[b78f59b61f1a8a0]::shared::Tzif<alloc[3370a9fffbfbb078]::string::String, jiff[b78f59b61f1a8a0]::shared::util::array_str::ArrayStr<30usize>, alloc[3370a9fffbfbb078]::vec::Vec<jiff[b78f59b61f1a8a0]::shared::TzifLocalTimeType>, alloc[3370a9fffbfbb078]::vec::Vec<i64>, alloc[3370a9fffbfbb078]::vec::Vec<jiff[b78f59b61f1a8a0]::shared::TzifDateTime>, alloc[3370a9fffbfbb078]::vec::Vec<jiff[b78f59b61f1a8a0]::shared::TzifDateTime>, alloc[3370a9fffbfbb078]::vec::Vec<jiff[b78f59b61f1a8a0]::shared::TzifTransitionInfo>>>::parse
       2278618 ┊    52.29% ┊ ... and 9446 more.
       2562533 ┊    58.81% ┊ Σ [9466 Total Rows]

The biggest offenders here from Jiff are the strftime and strptime parsing routines. But those don't show up at all for you. So from my perspective, it's very hard to make this issue actionable without something more concrete.

[curiousdannii]

I've built Glulxe three times, using Jiff 0.2.16, 0.2.24, and 0.2.33. This is from https://github.com/curiousdannii/emglken
The pinned versions in remglk/regmlk and remglk/remglk_capi are all that is changed between them.
Rust 1.94 and Emscripten 5.0.5 in case that matters.

This zip has the three .wasm files and their twiggy top results.
glulxe-jiff.zip

My code using Jiff is pretty much just these two blocks:

• https://github.com/curiousdannii/remglk-rs/blob/c459d64f3f66b2a1f683feb2b7c178f727597ffe/remglk/src/glkapi/mod.rs#L2040-L2089
• https://github.com/curiousdannii/remglk-rs/blob/c459d64f3f66b2a1f683feb2b7c178f727597ffe/remglk_capi/src/systems/emglken.rs#L153-L163

[BurntSushi]

Could you please provide the exact commands I can use to reproduce your results? If I can't do that, then I can't iterate. If I can't iterate, then it's very hard to help here.

So the reproduction should look something like this:

$ git clone ...
$ cd ...
$ cargo build ...
$ twiggy top -n20 ...

Or something of that variety.

Thanks.

[curiousdannii]

Sorry, yes of course!

git clone --recursive https://github.com/curiousdannii/emglken.git
cd emglken
head -27 CMakeLists.txt > CMakeLists.txt2 && cp CMakeLists.txt2 CMakeLists.txt # Just to skip the other 6 binaries
npm install
./build.sh
twiggy top build/glulxe.wasm

Needs Node (npm) and Docker. Emscripten doesn't need to be installed (it uses a Docker image.)