Chore

• v1.0.0
• be compatible with docs.rs See
• fix keywords
• see why makefile is not working
• found an illegal-instruction error Just run cargo test
• intermediate crate info from hunk However, it's not yet working
• first commit

Other

• docs for all remaining methods.
• documentation for crateversion
• customizations for us This could possibly work actually
• test osx as well
• allow to change seen-ref name
• show backtrace
• attempt of fetch_changes implementation It fails as it cannot create the correct reference.
• support for unyanking We just count it as adding a crate, which also makes sense.
• handle yanked files
• now seeing the first added crates
• automate running tests quickly Using an existing checkout is important enough to put it into a makefile for documentation and automation.
• support for commit'ishs for diffs
• first traversal method The test fails for the wrong reason though, as it fails to parse my refspec even though libgit2 seems to be able to do it properly, and a recent-enough version is used.
• test against all versions of rust
• simplify travis travis-cargo does nothing for us in this case.

New Features

• using openssl file as basis

Bug Fixes

• give up on osx Openssl can't be linked due to missing symbols, no matter what.

  Maybe ... the version is too new? Odd that it works locally though.

• see if openssl somehow works after all

• openssl setup from openssl-rs itself

• try the msvc 32bit version

• msys2 seems to be the problem here

• try to enforce linking latest openssl

• try disabling the cache Maybe that way, the 64 bit builds will work

• enforce more recent openssl lib Currently some symbols are missing or are named differently. Let's also see if it links it to the correct spot in /usr/local/opt/openssl

• only 1.1.0c seems to be working See

• remove 32 bit targets due to failure The previous version triggered this error: Error parsing appveyor.yml: Value can be either string or "secure" string. (Line: 5, Column: 7)

• attempt to set OPENSSL env vars

• try to allow 32 bit failures Could also remove them, but ideally it remains visible to maybe one day work.

• try without index-bare cache For now ... I have not seen it being cloned at all. Possibly the cache provides a none-empty directory ?

• custom code to get cmake onto osx

• let's hope it will actually run now

• set envvars propery

• remove debug prints

• remove error handling in makefile This isn't a good thing to do for all of those without cargo. However, let's see if it fixes the issue first.

• to work with GNU make

• finally the correct refspec Fair enough, need trees here after all!

• cmake binary installation

• use different cmake source

• cmake upgrade via before_script

• see if a custom source will get latest cmake

• enforce latest cmake

Refactor

• don't enforce branches Previously valid ref-names would be interpreted as branches, which limits their use despite the name indicating otherwise.

  Now every ref path is supported.

• use quick-error That way, the crate-error type is an actual error with all the relevant implementations.

  It's impossible to work without it, I think :).

• move crate into own module

• better diff.print implementation

• simplify fetch-changes error handling

• much nicer Crate from json It's good to know that we can chain everything, but that should only be done if there is an actual dependency of the respective values.

  Otherwise, just abort early.

• better error handling

• allow direct usage of objects as well

Test

• remove duplicate all crates are also iterated in the 'changed_since_last_fetch' test.
• fetch_changes() Just the frame and all testing I could think of
• assure we can handle the entire index
• first clone on demand Would be nice to be able to specify --depth as well.
• first failing test

Commit Statistics