git.shadowcat.co.uk Git - p5sagit/Try-Tiny.git/log

add Sub::Name and perl 5.010 to test recommendations (fixes #18)

skip test if Capture::Tiny is too old (Martin Popel, fixes #17)

Merge pull request #15 from Flimm/master

Add warning to POD about try/finally

Add warning to POD about try/finally

try/finally suppresses errors, which is unexpected behaviour if you come
from another programming language.

Merge pull request #11 from ilmari/global-destruction

Import 'import' instead of inheriting it

Merge pull request #13 from anaxagoras/master

Better docs on capturing success with catch block

second example now behaves the same as the first

Better docs on capturing success with catch block

While there is a note that calls it out below, I believe the behavior and usage of capturing success when also using a catch{} block is sufficiently different to the example that it merits its own example in the same code block. My rationale is twofold:

1) Because it is sufficiently different in behavior, giving an example goes a long way towards understanding.
2) Because this is a critical feature to have fail and it might be easy for someone reading the docs to digest the example and move on before reading the note below (which is unemphasized), it is best to have the second code example alongside the first with a meaningful name so that they are both digested at the same time.

Test loading during global destruction

Import 'import' instead of inheriting it

This fixes errors when the first use of Try::Tiny happens during global
destruction on Perl >= 5.10.0.

The problem is that the inheritance caches don't get invalidated during
global destruction, and base.pm skips classes that the calling class
already inherits from, thus populating an initial empty inheritance
cache, which doesn't get invalidated when it adds the class to @ISA.

Merge pull request #10 from pghmcfc/old-test-more

Avoid dependency on Test::More 0.88

The use of done_testing implies a dependency of Test::More 0.88 or
later. To avoid this, and provide compatibility with older versions,
remove done_testing and supply a test plan.

To ensure that the end of the test is reached (rather than exiting
early due to an error), I added an additional "pass" test to bump
the test count.

Merge pull request #9 from haarg/gd-fork-bug

work around perl bug where pads may be reused in cleanup after fork

work around perl bug where pads may be reused in cleanup after fork

During global destruction triggered by a explicit exit after forking,
perl may re-use the pad for try in a nested call in a DESTROY. All
variables used must be explictly initialized to avoid this.

remove accidental Sub::Name test dep

Merge pull request #8 from markstos/master

fix missing semicolon in docs.

Add missing semicolon in docs

This looked wrong without a semicolon, so I tested it. Sure enough, it was a syntax error without it, and ran as expected with it.

make this consistent

tests for finally name too

Change caller to show blocks belonging to our caller

name the try, catch and finally blocks

also die on catch/finally in scalar context (RT#81070)

update test counts for removing use_ok

update this link too

stop using use_ok, it's pointless

Merge pull request #7 from ribasushi/master

Various housekeeping etc

Fix linkage brainfart from 2245f1ae

The construct introduced in 2b0d579d8 confuses perl 5.6 - simplify

Make sure users are properly notified of (useless) exceptions in finally blocks

It is possible to rewrite Try::Tiny to invoke the finally blocks without
the use of scope guards, however to do this an extra eval frame around
catch will have to be introduced. Furthermore code in the wild may rely
on the "ignorance" of a finally() exception.

In short - the proper fix may no longer be feasible due to backcompat
issues, therefore the current warning seems like the only feasible
compromise.

correct prototypes in docs

Clarify exception on unexpected try() arguments

Sanity check erroneous invocations

Fix broken link to slides (RT#55659)

We already explicitly save the value of $@ - local()ization is not needed

511c05cae introduced an extension to try{} - unlike in an eval{} a try{}
block has access to the previous value of $@. Since we already have a copy,
we can do a manual restore of $@ instead of the expensive local() mechanism.

Results in a 7% speedup of bare try{} on 5.16

This is no longer a TODO, but given the experimental nature not ready to be documented

https://metacpan.org/module/SHAY/perl-5.17.3/pod/perldelta.pod#given-now-aliases-the-global-_

"Modernize" code a bit

Turn on warnings everywhere (5.6 being lowest target)
Silence experimental warnings in given/when tests

fix tests failing on 5.6.x due to differing DESTROY semantics

Detabardize - standardize code/tests on 2-space indent

Zero functional changes (examine under git diff -w)

changelog and version bump

this todo is only relevant before 5.14

MYMETA stuff to MANIFEST.SKIP

Merge pull request #5 from cebjyre/docs

Documentation fixes

Documentation fixes

Fixed a simple typo, and an internal POD link.

Merge pull request #4 from karenetheridge/topic/doc_20121114

Documentation improvements

C<> tags, minor punctuation and grammar fixes

add alternative mechanisms for fallback return values

add repo info to metadata

changelog and version bump

better example for returning from a try block (forwardever)

Test context/RV of try/catch/finally blocks

changelog and version bump

clarify the state of $_ in finally blocks

caveat about error-silencing

changelog and version bump

don't index ::ScopeGuard

changelog and version bump

add test for finally being called when catch dies

fix issue with blead (require 5.010 != use 5.010)

avoid a warning when testing with -w

no, do not sign our release

bump version for release

document changes for next version

document args to finally{} block

test for finally{} getting the error

allow multiple finally{} per try{}; pass error in

Add docs about @_ aliasing

use a plan instead of done_testing in TODO test

move use test to BEGIN block

add TODO test for lexical $_ gotcha