►
From YouTube: [Spanish] RustLatam 2019 - Esteban Kuber: Friendly Ferris: Developing Kind Compiler Errors
Description
The Rust Compiler is quickly becoming well known for its helpful messages, going out of its way to provide relevant suggestions to common problems.
As the Rust language reaches more people, the compiler is having to cater to both newcomers and highly experienced developers who may write incredibly complex code.
In turn, this means that diagnostics need to be comprehensive enough for newcomers without being distracting to those already familiar with the language. Making rustc act as a personalized tutor is useful for newcomers and has contributed to the language's growing popularity. However, it requires a lot of upfront work and foresight from the developers of the compiler.
During this talk we will cover the current guiding principles of rustc development, how you can contribute to the compiler, and where to find documentation and help.
Follow us on Twitter: https://twitter.com/rustlatamconf
A
A
Ago,
For,
some
time
I
had
been
hearing
about
a
very
low-level
programming
language
that
promised
to
allow
writing
very
efficient
code
without
memory
and
6,
but
without
the
need
to
have
a
collector
karma
and
in
May
2015
rast
was
released
to
the
world
and
when
it
gave
1
point
that
track
1.0,
I
said
it's
also
a
good
round
number.
It's
a
good
time
to
start
coming
back.
If
there
really
is
something
here,
if
the
promises
that
the
community
is
giving
really
have
meat,
it
wasn't
a
good
experience.
A
I
really
hit
a
wall
even
though
the
documentation
was
very
Well.
I
read
the
book.
The
concepts
really
made
sense.
It
was
from
a
theoretical
point
of
view.
It
really
made
a
lot
of
sense,
but
when
I
wanted
to
apply
the
concepts
I
had
to
hit
the
learning
curve,
which
was
a
wall,
and
it
was
really
a
little
frustrating
I
put
it
aside
for
few
months
until
in
the
holidays,
I
had
a
week
of
a
week
in
which
I
could
dedicate
myself
to
really
learning
the
language.
A
At
that
time
it
was.
The
experience
was
much
better.
In
the
meantime,
it
had
already
been
seven
months
of
improvements
to
the
compiler
improvements
to
the
ergonomics
of
the
language,
but
also
one
of
the
things
that
changed
is
that
I
had
a
particular
project
to
contribute
to.
Instead
of
just
writing
code
toys,
I
started
contributing
a
couple
of
pierce
and
debug
quest
array
be
that
of
an
autocompletion
library
which
now,
if
any
of
you
guys,
is
using
rast
and
uses
the
real
one
to
autocomplete
down
dozens
of
other
attachments
around
this
time.
A
It's
also
when
I
first
heard
about
fighting
with
the
drunk,
and
I
don't
remember
what
error
it
was,
but
I
remember
sitting
for
a
good
half
hour,
looking
my
screen
trying
to
figure
out
what
the
compiler
was
telling
me
and
when
I
finally
figured
out
what
the
problem
was,
I
didn't
feel
smart.
I
was
mad
that
the
compiler
wasn't
telling
me
with
the
problem
in
the
compiler's
mental
model.
The
compiler
knew
what
was
the
problem,
but
he
wasn't
telling
me
so.
A
He
starts
looking
at
the
ras
code
and
I
had
no
experience
with
any
real
world
compiler,
but
the
trace
was
written.
So
I
said:
ok,
let's
open
the
hood
and
see
what's
in
here
da
Let's,
see
if
I
can
change
how
the
compiler
is
talking
to
me
and
the
great
advantage
it
had
is
that
how
the
diagnostic
is
displayed
to
the
outside
was
something
I
could
use
greg
to
find
I
could
find
where
it
was
defined,
so
I
started
building
a
public
part,
it's
very
small
for
some
quality
of
life
improvements
for
development.
A
All
of
this
just
for
me
and
over
time
I
started
to
contribute
more
and
more
to
the
point
where
I
am
finally
a
part
of
the
team
and
of
everyone
here.
How
many
have
used
rares
even
a
little
bit?
How
many
of
you
have
had
the
experience
of
sitting
down
and
seeing
an
error
message
and
saying?
Ok,
interesting,
saves
me
a
lot
of
work.
Work.
Ok,
no,
not
all
I
hope
a
lot.
A
Ok,
something
that
is
visible
online
is
there
are
many
people
who
have
had
the
experience
of
having
is
to
be
surprised
by
the
error,
messages
that
it
is
like
this,
but
which
makes
me
very
proud
and
It
makes
me
very
happy
that
the
work
that
we
are
giving
is
not
really
visible,
but
at
the
same
time
I
put
many.
At
the
same
time,
there
are
many
people
who
have
also
found
the
other
part
of
this.
It
does
not
make
sense.
This
reminds
me
for
more
than
what
you
are
trying
to
tell
me.
A
A
That
is
not
so
good,
and
this
is
for
many
reasons,
some
if
you
are
doing
something
very
difficult,
that
the
compiler
has
no
way
of
mind
reading,
but
there
are
also
issues
that
we
haven't
gotten
around
to
working
on,
yet
that
you
would
have
to
affect
too
many
parts
of
the
inspirer
to
give
you
around
before
continuing
what
I
would
like
to
do
is
have
a
shared
vocabulary
for
those
who
have
already
used
rast,.
Surely
they
have
seen
an
error
similar
to
this
one,?
It
probably
is
not
there,,
but
they
have
seen
a
similar
error,.
A
It
is
not
counted
and
minutes,
well,.
It
would
be
a
bit
cut
off,,
but
these
messages
have
been
written
until
they
have
ever
seen.
A
all
error
messages
have
a
type
it
can
be
either
an
error
or
a
warning
in
warning,
and
as
and
can
optionally
have
a
code.
That
code
is
just
for
a
documentation
issue.
It's
to
allow
google
something
when
you
have
the
problem,
all
messages
errors
have
a
description.
That
description
explains
what
the
problem
is
in
a
general
way.
You
also
have
to
point
somewhere
it
points
to
who
filed
the
problem.
A
That
happens.
If
you
have
a
window
to
your
own
post
code,
where
the
compiler
points
to
which,
where
are
the
places
where
it
found
problems,
are
these
places
and
internally
we
refer
to
them
as
spanish.
The
each
spam
is
simply
a
a
location
in
the
code.
An
offside
of
bytes
and
they
can
be
of
two
types,
is
primary,
which
is
where
the
compiler
found
a
problem
and
children
where
the
compiler?
A
Would
you
want
look
home
and
there
are
also
all
the
bugs
can
be
extended
with
more
parts
that
look
good?
On
the
other
hand,
there
is
the
user
experience.
On
the
other
hand,
in
the
internal
part
of
the
compiler,
it
is
not
just
a
great
program
I
have
made
it
does
everything
al
At
the
same
time
is
when
the
compiler
is
given
a
small
piece
of
code
and
the
first
thing
the
compiler
does,
that
runs
alexander
alexeev
grabs.
A
The
text
file
bypass
by
byte
was
going
to
generate
tokens,
which
are
the
fundamental
units
of
language
is
just
every
%
of
the
code
is
observed
and
generates.
A
tree
of
all
this
that
is
showing
below
is
not
exactly
what
the
compiler
understands,
but
it
is
very
similar
having
that
touch
string,
we
pass
it
to
the
partner
who
does
not
see
this
the
tokens
and
what
it
generates
from
that
is
a
estiu
napster,
abstract,
syntax,
twitter.
A
The
major
definition,
but
if
you
look
closely
there
is
a
field
that
is
the
node
of
the
type
of
expressions.
This
is
all
that
the
compiler
understands
how
to
rush.
This
is
all
that
is
flush
is
all
in
front.
Are
all
the
valid
expressions
of
expression
in
this
step
when
the
compiler
is
given
the
code
on
the
left?
What
is
in
the
comments
does
not
subscribe.
A
What
are
the
expressions
that
are
seen
by
the
compiler
and
if
you
notice,
in
the
second
line,
I
just
put
a
literal
I
didn't
put
the
led
assignment
as
an
expression,
because
it's
not
an
expression,
it's
a
steak,
víctor
ros
statement.
They
have
to
end
with
a
semicolon
and
a
way
that
they
can
paint
it.
It
has
an
assignment.
It
is
an
expression.
It
is
with
the
error
you
give
and
they
forget
to
put
a
period
that
we
could
improve
and
the
compiler
could
explain
it.
A
The
next
step
is
name
resolution
and
macro
expansion.
These
are
the
literal
macros,
not
the
procedural
macros.
It
is
which
are
the
ones
that
you
were
seeing
with
davis
yesterday,
but
it
is
the
moment
in
which
you
check.
If
there
is
an
existing
method
for
this
variable,
the
next
step
is
the
transformation
of
the
stick.
A
The
tree
that
is
abstract
to
a
different
tree
that
has
information
more
oriented
to
rodas,
is
in
particular
of
information,
knows
that
any
blog
that
all
all
expressions
has
a
type
expression,
and
you
can
see
this
by
simply
calling
the
compiler
with
this
flag,
and
it
will
generate
the
same
code
that
you
gave
it
with
comments
of.
There
is
a
small
portion
of
the
information
that
the
compiler
has
after
this
step
is
the
type
check
with
stereotypes,.
A
The
compiler
simply
looks
at
all
the
expressions
and
checks
that
they
really
match,
and
if
they
don't
match
there
is
an
error,,
then
there
is
the
privacy
check,.
It's
just
like
that.
From
this
point,
on.
you
can
access
others
the
lifetime
check.
I
won't
translate
it
lifetime
analysis
doesn't
sound
good,
so
the
rest
I'm
going
to
use
sessions
times
in
this
step,
the
compiler
tries
to
see
and
make
sure
that
the
code
you
wrote
is
really
valid.
A
A
Do
what
are
the
four
fundamental
pillars
that
we
try
to
follow?
First
of
all,
we
want
to
have
bugs
that
make
sense.
They
need
perfect
generic
bugs.
We
try
to
give
a
error
that,
in
isolation
is
understandable
by
the
person
who
lifts
a
good
example
would
be
the
300
error,
300
other
different
ones.
You
gave
one
type
I
was
waiting
for
another.
This
is
a
very
common
error.
It
happens
all
the
time.
It
is
a
generic
error.
It
does
not
need
too
much
context.
A
This
code
is
this
to
generate
a
tracking
error
that
1.0,
and
if
you
look
it
is
a
bit
generic.
It
does
not
say
much
that
same
error
today
is
an
error.
300
others
is
a
type
error.
The
block
had
to
solve
there,
you
use
it
and
it
did
not
solve
because
it
did
not
have
a
mouse.
It
is
in
a
return
expression.
A
A
A
I
am
cheating
a
little
bit
because
it
is
harder
to
see
on
the
screen
like
this,
but
this
is
the
error
we
used
to
give,
and
this
is
the
error
we
give
now
what
was
happening
in
the
code.
Is
you
can't
have
two
variables
by
passing
two
variables
as
arguments
to
a
function
and
return
with
different
life
times
and
return
one
or
the
other,
defending
one
of
a
condition?
A
A
Is
this
one
that
is
still
not
perfect,
but
the
original
problem
is
that
the
trade,
the
king
function,
the
lights
in
the
trade
function,
are
different
from
the
implementation,
and
this
is
an
example
is
a
good
example
of
something
that
we
could
improve
and
that
we
have
improved,
but
that
we
could
improve
more
the
next
leg.
The
next
pillar
is
to
recover
from
error
states.
If
we
go
back
to
this
list,
each
of
these
steps
can
fail
for
reasons
different
genius.
A
It
is
very
common
if
they
have
been
written
for
some
time.
It
was
very.
It
was
very
common
to
write
a
piece
of
code.
You
had
a
syntax
error
fix
the
syntax
error.
The
name
of
the
variable
was
wrong,
fix
the
error
of
the
variable
that
was
correct,
return
to
compile,
and
there
it
tells
you
the
types
that
are
incorrect
well
solve
that.
And
finally,
you
give
yourself
the
lifetime
analysis.
It
gives
you
4000
errors,
and
you
realize
that
what
you
are
trying
to
do
cannot
really
be
represented
to
us.
A
That
is
a
very
good
bad
experience,
so
something
we
try
to
do
is
recover
from
error
states
of
industries,
that
is
to
say
recover
in
one
step
in
the
next
step,
assume
that
everything
was
fine.
To
give
an
example.
Here
is
a
function
that
I
am
defining
in
rast
as
external
that
can
be
used
in
cee,
and
I
'm
defining
it
as
bariatric,
but
various
functions
in
rast
can't
be
defined
can
only
be
consumed.
A
A
So
we
avoid
having
to
compile
many
times
until
finally
reaching
a
point
where
we
find
the
one
that
works,
and
the
next
one
step
is
to
hide
errors,
something
that
also
happens
when
we
started
by
allowing
incorrect
code
to
go
to
the
next
pro
phases,
and
there
were
errors,
there
was
a
cascading
effect
on
the
errors.
It
was
very
common
to
have
a
single
error
that
actually
generated
an
example
of
this
would
be.
A
This
is
a
table
of
three
elements
in
which
we
are
assigning,
but
we
are
assigning
a
dubious
pair
until
very
recently,
we
were
giving
this
error.
The
gus
are
four
errors.
Three
are
completely
useless
when
they
fix
the
first
three,
the
other
three
disappear.
So
now
what
we
do
is
hide
those
zeros
These,
four
things
that
we
do
are
precisely
to
try
to
improve
the
experience
of
each
user,.
But
there
is
a
fifth
thing
that
we
are
trying
to
achieve.
A
A
To
try
to
scale
there
are
two
things
we
try.
One
is
to
improve
the
ergonomics
of
the
language
itself
to
allow
things
like
non-lexical
life
times.
The
male
economics
just
accept
more
code,
and
the
other
side
of
that
coin
is
to
do
just
because
the
compiler
doesn't
tell
you
and
I
yelled,
but
rather
that
you
dress
up
I
'm,
trying
to
help
you
this.
You
can
see
there
are
other
people
who
also
think
in
the
same
way.
Luckily,
and
one
of
the
things
you
can
see,
there
is
something
I
heard
in
some
of
the
here.
A
This
provides
the
compiler
and
it
makes
six
months.
I
had
a
terrible
experience,
I
tried
them
now,
and
it
was
much
better,
and
luckily
that
is
something
that
shows
the
importance
of
the
work
we
have
been
doing.
We
have
the
compiler
not
only
give
you
errors,
but
also
suggestions
on
how
to
help
like
other
javier.
How
correct
your
mistake
here?
We
don't
want
everything
to
be
at
stake,
say:
ok,
it
may
be
working.
No,
we
don't
want
the
compiler
of
this
yelling,
it's
all
wrong.
A
To
give
you
an
idea
of
the
work
we've
done
for
this
code.
These
are
the
errors
of
1.0.
Until
today,
it
is
a
very
simple
error.
It
is
a
type
error,
but
this
is
how
it
would
look
in
1.0
and
over
time
you
can
see
that
we
were
integrating
to
try
to
achieve
changes,
sometimes
subtle,
sometimes
very
small,
but
that
help
to
hitler.
The
message
in
general
be
dignified.
A
A
Very
welcoming
the
compiler-
and
this
is
a
universe
that
before
people
like
any
of
you
here,
what
I
would
like
you
to
do
when
you
leave
here
when
you
see
an
error
that
you
do
not
understand,
write,
write,
documentation,
raise
a
ticket,
I
have
left.
If
you
have
the
time
contribute,
the
compiler
is
just
another
one.
As
you
can
see,
it's
a
big
application,
but
it's
an
application
itself.
It's
not
magic.
We
do
n't
have
code,
we
have
a
binary
in
the
middle
magic
is
something
that
any
of
you
can
start
contributing.
A
I
started
contributing
to
rast
less
than
six
months
later,
having
started
to
learn
the
language
and
I
really
learned
rast
with
the
compiler.
The
community
is
going
to
help
us.
There
are
many
resources.
The
rast
forge
trust
force
force
for
the
road.
Langdon
org
is
a
very
important
resource.
It
has
many
links.
The
information
has
information
about
there,
a
book
about
the
compiler
that
will
allow
you
to
navigate
the
code.
You
can
see
what
the
process
of
contributing
is
like.
There
are
a
lot
of
resources
to
be
able
to
contribute
to
the
compiler.
A
There
are
also
forums.
There
are
different,
chats
that
you
can
use
either
yesterday
or
if
you
can
use
scioli
can
go
to
disks.
There
are
always
people
from
the
teams
available
to
help,
also,
even
if
they
have
some
use
case,
that
they
do
not
see
represented
but
make
a
I
lean
back
as
a
proposal
for
a
change
to
the
language,
either
from
the
economy
of
the
language
itself
from
how
it
works
internally,
you
can
do
it.
A
Any
of
you
can
do
it.
I
didn't
have
much
I
didn't
have
much
experience,
and
it
really
is
very
satisfying,
and
this
making
the
act
like
a
tutor
sometimes
requires
a
lot
of
work,
but
it
is
very
important,
and
not
only
for
the
experience
of
people
who
are
already
using
trace.
It
is
very
important
for
people
who
are
going
to
use
trace
a
programming
language
that
does
not
have
a
ramp
to
attract
new
developers
and
it's
a
language.
A
That's
going
to
I,
don't
want
to
use
the
word
die
because
it's
too
strong,
but
it's
not
going
to
grow.
It's
super
important
and
there
are
a
lot
of
languages.
that
are
now
doing
this
kind
of
this
kind
of
thing
and
I
really
want
it
to
rast.
I
still
have
the
cases
where
the
good
mistakes
happen.
More
often,
if
the
cases
where
the
bad
mistakes
happen
less
often
and
since
I
obviously
speak
faster
than
every
time,
I
practice
I.