Russ Cox [Sun, 17 Oct 2010 15:41:23 +0000 (11:41 -0400)]
5l: stop using R12 as SB
Because the SB is only good for 8k and Go programs
tend to have much more data than that, SB doesn't
save very much. A fmt.Printf-based hello world program
has 360 kB text segment. Removing SB makes the text
500 bytes (0.14%) longer.
Russ Cox [Fri, 15 Oct 2010 19:18:47 +0000 (15:18 -0400)]
6l: function at a time code layout
Also change the span-dependent jump algorithm
to use fewer iterations:
* resolve forward jumps at their targets (comefrom list)
* mark jumps as small or big and only do small->big
* record whether a jump failed to be encodable
These changes mean that a function with only small
jumps can be laid out in a single iteration, and the
vast majority of functions take just two iterations.
I was seeing a maximum of 5 iterations before; the
max now is 3 and there are fewer that get even that far.
Russ Cox [Fri, 15 Oct 2010 03:48:40 +0000 (23:48 -0400)]
5l, 6l, 8l: accumulate data image during import
Using explicit relocations internally, we can
represent the data for a particular symbol as
an initialized block of memory instead of a
linked list of ADATA instructions. The real
goal here is to be able to hand off some of the
relocations to the dynamic linker when interacting
with system libraries, but a pleasant side effect is
that the memory image is much more compact
than the ADATA list, so the linkers use less memory.
Russ Cox [Wed, 13 Oct 2010 20:20:22 +0000 (16:20 -0400)]
various: avoid %ld etc
The Plan 9 tools assume that long is 32 bits.
We converted all instances of long to int32 when
importing the code but missed the print formats.
Because int32 is always int on the compilers we use,
it is never correct to use %lux, %ld, etc. Convert to %ux, %d, etc.
(It matters because on 64-bit gcc, long is 64 bits,
so we were printing 32-bit quantities with 64-bit formats.)
Russ Cox [Wed, 13 Oct 2010 19:51:21 +0000 (15:51 -0400)]
5l, 6l, 8l: first pass cleanup
* Maintain Sym* list for text with individual
prog lists instead of using one huge list and
overloading p->pcond.
* Comment what each file is for.
* Move some output code from span.c to asm.c.
* Move profiling into prof.c, symbol table into symtab.c.
* Move mkfwd to ld/lib.c.
* Throw away dhog dynamic loading code.
* Throw away Alef become.
* Fix printing of WORD instructions in 5l -a.
Goal here is to be able to handle each piece of text or data
as a separate piece, both to make it easier to load the
occasional .o file and also to make it possible to split the
work across multiple threads.
Jim McGrath [Tue, 12 Oct 2010 20:52:17 +0000 (16:52 -0400)]
6l: work with OS X nm/otool
6l was skipping emitting the (2 byte) symbol table if there were no imported or exported
symbols. You can't just drop the symbol table entirely - the linker dies if you have
a linkedit section but no table. You can omit the linkedit section or both the linkedit
and the dlyd parts in the right circumstances, but that seems much more risky to me.
Rob Pike [Tue, 12 Oct 2010 19:59:18 +0000 (12:59 -0700)]
log: new interface
New logging interface simplifies and generalizes.
1) Loggers now have only one output.
2) log.Stdout, Stderr, Crash and friends are gone.
Logging is now always to standard error by default.
3) log.Panic* replaces log.Crash*.
4) Exiting and panicking are not part of the logger's state; instead
the functions Exit* and Panic* simply call Exit or panic after
printing.
5) There is now one 'standard logger'. Instead of calling Stderr,
use Print etc. There are now triples, by analogy with fmt:
Print, Println, Printf
What was log.Stderr is now best represented by log.Println,
since there are now separate Print and Println functions
(and methods).
6) New functions SetOutput, SetFlags, and SetPrefix allow global
editing of the standard logger's properties. This is new
functionality. For instance, one can call
log.SetFlags(log.Lshortfile|log.Ltime|log.Lmicroseconds)
to get all logging output to show file name, line number, and
time stamp.
In short, for most purposes
log.Stderr -> log.Println or log.Print
log.Stderrf -> log.Printf
log.Crash -> log.Panicln or log.Panic
log.Crashf -> log.Panicf
log.Exit -> log.Exitln or log.Exit
log.Exitf -> log.Exitf (no change)
This has a slight breakage: since loggers now write only to one
output, existing calls to log.New() need to delete the second argument.
Also, custom loggers with exit or panic properties will need to be
reworked.
All package code updated to new interface.
The test has been reworked somewhat.
The old interface will be removed after the new release.
For now, its elements are marked 'deprecated' in their comments.
Russ Cox [Tue, 12 Oct 2010 03:58:51 +0000 (23:58 -0400)]
arm: fix build
Effectively reverts https://code.google.com/p/go/source/detail?r=8c52477401ad
Should make ARM build pass again, but untested.
Probably still bugs involving reflect.call somewhere.
Russ Cox [Tue, 12 Oct 2010 02:38:42 +0000 (22:38 -0400)]
exp/iterable: delete
Package iterable has outlived its utility.
It is an interesting demonstration, but it encourages
people to use iteration over channels where simple
iteration over array indices or a linked list would be
cheaper, simpler, and have fewer races.
R=dsymonds, r
CC=golang-dev
https://golang.org/cl/2436041
Rob Pike [Mon, 11 Oct 2010 19:40:13 +0000 (12:40 -0700)]
new command gotry.
An exercise in reflection and an unusual tool.
From the usage message:
usage: gotry [packagedirectory] expression ...
Given one expression, gotry attempts to evaluate that expression.
Given multiple expressions, gotry treats them as a list of arguments
and result values and attempts to find a function in the package
that, given the first few expressions as arguments, evaluates to
the remaining expressions as results. If the first expression has
methods, it will also search for applicable methods.
If there are multiple expressions, a package directory must be
specified. If there is a package argument, the expressions are
evaluated in an environment that includes
import . "packagedirectory"
Russ Cox [Fri, 1 Oct 2010 20:02:18 +0000 (16:02 -0400)]
Make.pkg: remove .so before installing new one
On Linux, overwriting an mmap'ed file causes
all the MAP_PRIVATE pages to get refreshed
with the new content, even ones that have been
modified by the process that did the mmap.
One specific instance of this is that after the
dynamic linker has relocated a page from a .so,
overwriting the .so will un-relocate it, making
the next use of one of the no-longer-relocated
addresses incorrect and probably crash the
program.
Linux must go out of its way to break programs
in this way: the pages have already been copied
on write, so they're not shared with the file system
cache, and it trashes them anyway. The manual
says the behavior when the file gets overwritten
is "undefined". Removing before copy avoids the
undefined behavior.