CVS log for parser3/src/main/execute.C

Up to [parser3project] / parser3 / src / main

Request diff between arbitrary revisions - Display revisions graphically

Keyword substitution: kv
Default branch: MAIN

Copyright year updated, websites links changed to https://

removes irrelevant lines from exception stack trace when exception occur in array assigment (related to bug #1165)
+ useless #define removed

feature #1273 core implementation: value::as_wint added, implemented for int/double/string, and used where applicable; vint now holds pa_wint type; format_double moved to pa_int

bugfix: when method is attached from another class, it's  junction_template is initialized with another class, thus get_vjunction should be used (related to feature #1286)

fixes for --with-build-warnings

Copyright year updated

absent constructor message fixed

Temp_recursion class added;
spelling: $LIMITS.max_recoursion -> $LIMITS.max_recursion

minor error messages spelling fixes

OP_CONSTRUCT_ARRAY added to support $a[v1;v2;...] syntax (related to feature #930)

optimization: Value::as() removed as dynamic_cast is faster, Value::is() is still required for is operator.

Array::Iterator unified with Hash::Iterator, has_next() replaced with operator bool (related to feature #930)

Warning war: clang changed -Wdeprecated-register warning to -Wregister, it's easier to remove all register vars then continue this fight. :)

optional @auto[] inheritance implemented if defined with two args (path, class name) (implements feature #1233)

volatile added to sync behavour with other OP_NUM_*, related to bug #1230

force double operands to be converted to double to avoid 80 bit operations in x87 mode (fixes bug #1230)

copyright, authors actualized

get_class_ref added and used

Copyright year updated

r.absolute -> r.full_disk_path

better error reporting when $result[^hash::create[]] in @main.

frame.result() should be used + minor optimization

execute_*_method simplified and unified

optimization: PUT_ELEMENT_REPLACED_ELEMENT no longer used in object-prototype mode

removes irrelevant lines from exception stack trace when exceptions occurs in assigment (fixes bug #1165)

Copyright year: 2015->2017

^reflection:stack[] added (implements feature #1052)

Request::SKIP_INTERRUPTED, Request::SKIP_RETURN added and implemented,
Request::SKIP_* now checked after each possible process/execute call (fixes bug #1077)

method call exceptions improved as method now keeps its name

Request::construct exception optimized

VS warning war: C4239: nonstandard extension used : 'return' : conversion from 'Value' to 'Value &' (related to feature #1020)

VS warning war

filename -> filespec, ^reflection:filename[$method] added (related to feature #1053)

Optimization: VMethodFrame now divided into VNativeMethodFrame, VParserMethodFrame, VLocalParserMethodFrame;
METHOD_FRAME_ACTION, EXPRESSION_FRAME_ACTION, CONSTRUCTOR_FRAME_ACTION defines added and used (related to feature #1104)

cleanup: r.write_pass_lang, r.write_no_lang -> r.write (implements feature #1091)

cleanup (related to feature #1104)

VExpressionFrame added, in_expression removed (implements feature #1104)

optimization: write_assign_lang removed as not required (first part of feature #1091 implementation)

opcode OP_PREPARE_TO_EXPRESSION removed as not required, optimized WContext is next task (implements feature #1104)

compilation without OPTIMIZE_CALL fixed

*** empty log message ***

optimization: intercept_string removed (implements feature #1097)

proces_to_value -> process after StringOrValue removed (related to feature #1020)

VALUE__GET_BASE_CLASS added to opcode_names

whitespace

optimization: StringOrValue removed as it just slows down and adds complexity (implements feature #1020)

request.classes(): Value -> VStateless_class, part 2 (related to issue #1051)

no more somebody_entered_some_class, $BASE:property works again, ^BASE:method[] bugs fixed (fixes bug #1059)

CLASS_SYMBOL, CLASS_NAME_SYMBOL added and used (related to feature #1026)

returned Value in .CLASS case for VJunction (related to feature #1051)

classes hash now contains only VStateless_class derived objects, .CLASS is also allways VStateless_class derived object (related to feature #1051).

name() removed from VStateless_class, type() is used instead (related to feature #1051)

Copyright year updated

GET_ELEMENT4CALL added to get_method_filename (related to feature #1017)

get_element4call implemented to distinguish ^hash.method from $hash.field (implements feature #1017)

optimization in getter-junction processing (related to issue #997)

process_getter result now checked if junction returned (closes issue #997)

removed most of the clang compiler warnings (closes issue #1018)

- $var.CLASS & $var.CLASS_NAME now have special opcodes. This can be switched off by commenting #define OPTIMIZE_BYTECODE_GET_ELEMENT__SPECIAL on pa_opcode.h ( new feature: #844 )

optimization: put_element no longer has "bool areplace" argument, related to issue #903

result should be pushed after VMethodFrame destructor is called, as it deletes junctions from stack params (fixes issue #868)

ident now works under Linux + ident displays filenames (closes issue #818)

ident now works under Linux + ident displays filenames (closes issue #818)
Copyright updated

void now inherited from string (feature #111)

constructor returning another object feature returned

cosmetic optimization in request:execute_method usage

default setter support + anti-recursive default getter support ( new feature: #13 )

optimization:
VMethodFrame(junction, caller) replaced with VMethodFrame(method, caller, self)
op_call(VMethodFrame &frame, bool constructing) removed, construct(class,method) added

- get_method_filename was moved to execute.C

bugfix: #59 OPTIMIZE_BYTECODE_GET_CLASS macrodefinition removed

fixes #60 Stacktrace now contains info for OP_CONSTRUCT_OBJECT and OP_GET_CLASS

- core in $table1.$table2 was fixed

- autouse was implemented

- optimization

- problems with properties in new classes implementation were fixed

- now constructor shouldn't be defined in class. it could be defined in parent.

- new classes implementation

- OPTIMIZE_BYTECODE_CONSTRUCT_OBJECT and OPTIMIZE_CONSTRUCT_OBJECT can't be disabled with defines any longer

no message

- value param was removed from WContext::WContext

- less calls for get_constructing()
- little refactoring

- create object for storing object's fields only for parser objects

- bytecode optimization of ^class:constructor
- optimization of single string write

- more optimizations for WITH_SELF, WITH_READ and WITH_ROOT
- little refactoring

- little refactoring
- optimizing write_assign_lang in taint/untaint

- optimizations for $self.field, ^self.method, $self.field[...] and $self.field(...) were added

- refactoring in last optimizations ($a[...], $a(...), $.a[...], $.a(...)). a bit less optimizations now but code is much easy and it is ready for further improvements.

- changed in destructing junctions

- core while printing in debug log non-printable values was fixed
- $a[$b.c] and $a[$b.$c] optimisations were added

no message

- call constr optimisation under separate define now

- bugfix: $a[] $b[$a] -- in $b should be string, not void

- optimisations for $a[^b[...]], $.a[^b[...]], $a(^b[...]) and $.a(^b[...])

- bytecode optimization for $a(1), $.b(2), $c[d] and $.e[f] was added

- debug printing moved in defines

- $a.$b & ^a.$b bytecode optimisation

- code for handle new opcodes OP_GET_ELEMENT_FIELD and OP_GET_ELEMENT_FIELD__WRITE was added

- result optimisation
- helper_length parameter removed from string constructors

- more optimisations were added: some operators don't switch write context anylonger

- simplifying process for getters

- changes in stack impl
- get_element() optimisation: 3rd param removed
- bugfix: problems with reading empty input param
- defines SAVE_CONTEXT and RESTORE_CONTEXT were added and used

- we can work without opcode OP_STORE_PARAM, so it was removed

- op-codes optimisation:
  1. VALUE+GET_CLASS=>VALUE_GET_CLASS
  2. WITH_READ+VALUE+GET_ELEMENT=>VALUE__GET_ELEMENT (not all yet)
  3. WITH_READ+VALUE+GET_ELEMENT__WRITE=>VALUE__GET_ELEMENT__WRITE
  4. WITH_READ+VALUE+GET_ELEMENT_OR_OPERATOR=>VALUE__GET_ELEMENT_OR_OPERATOR

- more replacements '*new VBool' to 'VBool::get'
- checks for ^break[], ^continue[] and parser.interrupted were optimised

- junction-optimisation (destructors)

- some junctions will be cached now

- VVoid::get() => new VVoid (one void-instance)

- bugfix: opcodes must be in separate namespace while debug execution as well

- opcodes were moved to separate namespace

- go back to one VMethodFrame with internal switch between local/global vars

- default getter soul

- code of OP_CALL and OP_CALL__WRITE was moved to separate method op_code
- switch from VMethodFrame to VMethodFrameGlobal + VMethodFrameLocal

- code frame don't intercept strings any longer

- little code reformating

- "parser.runtime" strings were removed

+ ^break[] ^continue[], in ^for, ^while, ^menu, ^hash/hashfile.foreach

! switched to VS2005 (projects and object<info explicitly typed>.for_each(, info))
+ $response:field[] setting void means removing
+ grammar $name` name stops now, for mysql `$field`
+ started ^break[] ^continue[], in ^for. TODO: to other iterators (while, menu, foreach)
+ all log messages += [uri=, method=, cl=]

merged from HEAD:
! changed file/dir_readable to simple file/dir_exist, this would help in situations "class not found because .p file has bad rights" << in that case error would be explicit "access denied to 'this' file"

! changed file/dir_readable to simple file/dir_exist, this would help in situations "class not found because .p file has bad rights" << in that case error would be explicit "access denied to 'this' file"

2005

. merged VJunction(new Junction) optimization from 3.1.4

+ optimized new VJunction(new Junction(params)) to new VJunction(params), thousands of mallocs removed

. 2005 ;)

! bug fix: static parent fields were not replaced in case $derived:field[put]
+ property setter can now be overridden

+ 3.2.0 beta: merged from dynamic_fields_join

! restored put_element('replace' param), and restored its checks in vobject.put_element->static fields | dynamic properties

joined_dynamic_fields: removed unused param from put_method (returned to 3 params)

joined_dynamic_fields: just compiled

joined_dynamic_fields: started

. properties: set works [alpha2]. for classes. todo: for objects

. properties: set works [alpha1]

+ started property [getters work but think of changing them too to precaching]

beauty: simplified system exception handling

attempt_check_call_in_explicit_result_mode: failed

2004

no message

old forgotten todo: when can't report problem (undefined @unhandled_exception) problem source string were not reported

static const char * const IDENT

turned on warnings level4 on all projects (except libltdl=off & libsdbm=level3)
found several unitialized vars

resurrected: pool idea. now only for destructing objects at request processing end

bugfix: ^throw context were saved/restored incompletely
http://www.parser.ru/forum/?id=21484

@postprocess  now takes $response:body/download

moved tempate_gc to HEAD

stack top_index() fixed

n-th attempt to make a=a work with double.
problem: as_double returns it's result in fp-register
compiler optimizes access to that register after b->as_double,
and just compares ALREADY_STORED_VALUE with prev-calculated a->as_double from memory.
_SAME_ double values do not match here.
when forced to REload fp-register, values do match.

tried to make them volatile.

n-th attempt to make a=a work with double.
problem: as_double returns it's result in fp-register
compiler optimizes access to that register after b->as_double,
and just compares ALREADY_STORED_VALUE with prev-calculated a->as_double from memory.
_SAME_ double values do not match here.
when forced to REload fp-register, values do match.

tried to make them volatile.

no message

debug info format simplified, now it's:
OP_VALUE
Operation::Origin << here
value*

higher limits:
  file number (max: 255)
  line number (max: 64535)
  column number (max: 255)

MethodParams& now [methods without params receive zero reference, but they are expected not to look there]

slightly improved error messages text:
1. is '%s', it
2. method undefined in case of ^void[]

exception handling done. now we have error columns
todo: test it

strack trace reset after handled exception [old bug fixed]

debug info: started using.
todo:complete

prepared debug info
todo:use it

gif encoder bug fix [gdGrowingBuf]

more locally scoped vars in execute (more easily optimizable)

execution stack copied to local
register
variable, removed lots [~2e7) of memory accesses

a number of bugfixes [while testing first real site (aval)]

hashcode implemented fully, including per-char callback [it can be - substr CORD node]

attempt to do gif encoding to CORD_ec [bad]
would rewrite as realloc now

classes.lib
main.lib compiled

pa_request.C 50% compiled

execute.C compiled

introducing StringBody [C++ CORD wrapper]

untaint.C 30%,
pa_common.C [done, without http for now]

started test

started porting to gc:
PA_Object done
lots of replacements also

gc logging

libgc attempt

set_base, set_derived simplified [counter leaks fixed]

SIGPIPE now can be intercepted
and does not cause exception in exception handler

xml: test: 107 bug fixes

tests:
101 method/variable name conflict test
102 form fields change should not not change anything [bug fix]

tests:
61 file::stat size
62 string->int autoconvert
63 double .int,dec,div,mul
64 date compare
65 file: basename,justname,justext
66 math:crypt
67 string.match simple on long
68 string.match normalized simple on long
69 string.int
70 file::stat content-type
71 table.join
72 hash parameter conditional pass
73 date daylightsaving,yearday
74 date arithmetics
75 response date values/attributes
76 string.pos of void, void.pos
77 syntax parsing
78 cookie tainting
79 switch with local
80 scientific numeric literal in string autoconvert
81 xor: logical and numerical
82 for delims
83 menu delims
84 table.hash distinct
85 long string  replace
86 throw+catch current language preserve
87 bit shifts
88 junction tests + $caller test
89 hash.foreach selfmodification
90 int/void to int
91 $caller test
92 junction is + def junction tests
93 $caller test
94 syntax test
95 table.hash(keygenerator)  [bug fixed]

test:
59 table.locate [bug fixed]
60 string eq string [bug fixed]

uchar changed to String_UL in all places [was not everywhere]
untaint test: 032

fixed bug with name_cstr mutable CharPtr
more tests

few bad get_string(0) [should have been get_string(&pool)]
todo: check other such calls

test10, bad lookups

xxx* xxx=smartptr.get(); << is bad style, after ";" original object got destructed, and xxx points to sky

release mode now compiles OK
test to show benefits of free

no message

another bug of unitialized simple-typed field fixed.
todo: find&kill all like that one

few bugs in debug output fixed

stack-stored items need extra .ref to prevent object_ptr from delete[them]

stackItem fixed [there were no string on stack before]

first letters out OK

*** empty log message ***

yuk: gcc on cygwin yelds "virtual memory exhausted" while trying to compile execute.C [eating up to 127MB]

main.lib - 0 error(s), 0 warning(s)

execute compiled

started last file from main library: execute

most pa_request compiled

grammar compiled

value compiled

started main: compile_tools

more compiled.
struck with string::store_to needing to know source/client charsets

2002->2003

$image.exif support
$image.exif.DateTime & co

SIGUSR1 >> writes to error_log uri of currently processed document
SIGPIPE >> interrupts request processing [exception = DB rollback]

comment

found out why, checked that for now.
todo: find out a way for that user could do that

static fix merged

static call $self fixed

no message

comment

$.name outside of $name[...]
checked

Request::self considered equal to VMethodFrame::self, and removed,
Request::get_self() mapped to VMethodFrame.self()

Request::self considered equal to VMethodFrame::self, and removed,
Request::get_self() mapped to VMethodFrame.self()

removed last pieces of old code allowing $junction.xxx at compile time

compiling to system classes disabled

removed $junction.get $junction.set[]
introducing $caller

operators now main-class-methods

operators @auto now executed in MAIN context

operator execution context now = MAIN, not closest stack frame

@touchit[]
$i[after]

----t.html
$i[before] << local
^touchit[]
$i << now 'before', were 'after'

---t.html
$i[before] << notlocal [main]
^touchit[]
$i << now 'after'

comment

no message

code junction calls disabled [before: code was compiled in such a way, that there were no code-junctions in OP_CALL]

this now error:
@badjunctioncall[]
^badjunctioncallinside{code}

@badjunctioncallinside[code]
^code[]

introducing $junction.get/put(1)
one can write iterators now:

^user-foreach[key;value]{$key=$value<br>}

@user-foreach[key;value;code]
^for[i](1;10){
	$code.key($i)
	$code.value($i*2)
	$code
}

<< >> int shifts

removed changes, operators executed with calling self.
lots of code with ^include code relies on defined/defining self variables

$a[1]
^include[print_a.p]
print_a.p: $a

^include[set_a.p]
a=$a
set_a.p:
$a[1]

operators now executed with MAIN self.
it's for ^include sake, too strong a change

found ancient param to Junction, removed.
allowed passing self to native_code_operators [for ^process to work, she needs self]

found ancient param to Junction, removed.
allowed passing self to native_code_operators [for ^process to work, she needs self]

allowed Request.self to be 0, checked that in VMethodFrame get/put and $self.

VCodeFrame parent param were specified badly

WWrapper which used in constructing objects(second param to sql method) ^...sql{}[$.default{code}]
now has parent, wich helps code in hash to survivi

removed redundant param to execute [stack said 'thanks']

moved junction kill responsibility to wcontext

double:sql badly called write_assign_lang, not write_no_lang,
thus doing unnecessary double/string converstion, which were loosing time&precesion

removed double_result, didn't help

no message

double_result made to move that var away from hungry g++ optimizer (-O2),
before: it were optimized and comparison operators worked badly

no message

no message

fixed numeric < & co so that thay now use c=a-b, c OPERATOR 0.
this works fine on solaris/intel for still unknown reason.

junction_cleaner moved to auto VMethodFrame [called less frequent, allowed to remove ugly Junction.change_context-s from many places, switch, mail:send..html{}, ..]
request.root renamed to method_frame [more easyreading]
ancient {...PUSH/POPs...} changed to stack vars [speed up]

fixed virtual calls

instead of type() checking everywhere used Value.as
now user descendants can be used in params.

VObject::as_*, is_defined now taken from bases.
xtable(table)
^if($xtable) now OK

is->as

is now works with VObject & VClass

introduced Value::get_element(..., bool looking_down) [needed to exclude endless recoursion]

VObject.get/out now looking down/up tree
todo: fix error message on non-constructed parents

^base:create[] dynamic call rewritten
todo: thorough testing

^base:create[] dynamic call rewritten
todo: thorough testing

killed VAliased [redundant], moved $CLASS to VObject only [parser class instance]

started as()

error reporting on object writes to MAIN improved [were ruind with fixing $obj[^if(1){$obj}] ]

vcodeframe were mistakenly not completely transparent to object writes.
$hash[^if(1){$hash}] now works

$Date: now$

ident.C* removed

marged latest HEAD updates, mainly file:justname&co & @auto/conf[filespec

introducing @conf/auto[filespec]

@rootconf [were @conf]

rolled back to before_killing_userjunction_contexts

user junctions context killed

junctions to local contexts got cleanized&checked later

OP_GET_METHOD_FRAME merged with OP_CALL, VCodeFrame move to stack [no more heap waste on each CALL]

removed Value::fname

$result in @main now taken into account as in usual functions

$result in @postprocess & @unhandled_exception now taken into account as in usual functions

no message

OP_CALL -> OP_CALL__WRITE and used that for removing VString wrapper

fixed name update

started dual write_xxx_lang functions, if checked & works

CodeFrame soul fixed [mistakenly killed by prev changes]

StringOrValue wcontext result, now ready for dual writes

OP_GET_ELEMENT+OP_GET_ELEMENT__WRITE
changed to OP_WRITE_VALUE
in var get cases

switch in hash constructor fixed [were problems with using of stacked wwrapper after it's death]

Request::process_internal codeFrame & wwrapper stacked [not wasting  heap anymore]

killed Request::process() wrapping VString(String) in case we need only String

assigned exception types

decided agains resetting exception_trace, just changed name

//#define DEBUG_STRING_APPENDS_VS_EXPANDS

strign to string append optimiziation idea and estimates

in_expression
aimed to solve old problem with string/nonstring values,
now in expressions double/int values are passed as-is, without stupid tostring/fromstring conversions

name spelling

2002

made operators lookup first in ^xxx situation.
so that one could not do
$if[1]
^if[xxx]

operators check 6.1
rewritten. now check is done at compile time

operators check 5
floated up old error of staying in 'entered class/object' state after $a($a..)
it prevented operators from being called thereafter

operators check 4
floated up old error of staying in 'entered object' state after ^a.a(a)
it prevented operators from being called thereafter

operators check 3
floated up old error of staying in 'entered class' state after $a:a(a)
it prevented operators from being called thereafter

operators check 2
floated up old error of staying in 'entered class' state after $a:a[a]
it prevented operators from being called thereafter

operators check, thay mistakenly seen in
  $class:operator
  $object.operators
contexts

merged from r17 - hashfile without logfiles

partial logical && || evaluation

$result in @main @postprocess @exception

junction evaluation canceled - endless recursion detected

die in pool::fail

exception with dynamic buffer re-back
now with counter

merged with no_bad_constructors_try

fixed this:
$bred[^string::length[]]
$bred[^response::clear[]]
$bred[^int::int[]]

now VObject creation moved into overriden VClass::create_new_value, thus 0 from create_new_value means that class has no ctors

$status:db
fixed used not initialized in db_connection & db_table

changed root behavior in native calls.
it left unchanged, so ^for ^foreach & co can use r.root to write their i & key,value there

\ in expression: int divide 10/3=3

debugging memory
WContext
flags to bits

total 5 652 652/158731
total 5 469 332/158731

debuggging memory.
string::as_int/double now usually uses stack, not heap [200K out of 17M :)]
VString(String&) not copies reference [29M -> 17M]

hashfile -100

stack backtrace

switched to c++ exceptions 0

z

vcodeframe made transparent enough to handle hash if creation&passing

noticed that vcodeframe not transparent enough, would change now

added ident.awk and main.dsp splitted to main+classes+types along to .am

merged latest bugfixes from before_xml branch
^if(def $hash) now true only when ^hash:_count[]!=0
moved
.string .save .file from dom to dnode

operators self changed to root
^for variable context changed to self

expressions unary+
^hash.count[]
bugfix: in expression compound($aa.zz.xx) names now can have minus '-' in name after '.'

introducing Array_iter

detected quick_get recursion bug. changed to get@execute, but needs array iterator, separate from array to avoid it & use caching

found another multithread bug in op.C (last, last global var killed) :)

hash creation syntax problem persists. restored $: for a while.
thinking of $man[$.age[zzz]] syntax now

fixing :: realization #1

introducing ::
this is constructor call prefix.
ordinary : remains for static accesses

$var[] is now empty string, not void

match replace code context moved to implicit $match context

with killed, code storage introduced with former 'with' syntax

first get_element, next get operator

only ^class:method dynamic calls allowed.
^BASE.method call disabled.
BASE element globally removed

$ORIGINS(1) output tracing mode

auto.p[@auto], /news/auto.p[no @auto], so that initializing second would not call first @auto

auto.p[@auto], /news/auto.p[no @auto], so that initializing second would not call first @auto

auto.p[@auto], /news/auto.p[no @auto], so that initializing second would not call first @auto

order of MAIN parents was wrong, fixed

math 0

 sin asin cos acos tan atan

/0  %0 checkes ver 2

/0  %0 checkes ver 2

/0  %0 checke

*/
static const char *RCSId="$Id: double.C,v 1.31 2001/06/28 07:41:59 parser Exp $";

static char *RCSId="$Id$";

nothing renamed to void

operators are not in root class again

'unknown' renamed to 'nothing'

'unknown' renamed to 'nothing'

hash:sql moved to main trunc. operators.txt updated

0

z

#include "pa_config_includes.h"
removed from most .C

endless recursion line no

ANTI_ENDLESS_EXECUTE_RECOURSION

removed /*SYNCHRONIZED*/  from hash.h

array debugged; adjusted

numbered params had wrong name - for instance: bad error message in ^for[] bad body type. fixed

op: MAX_LOOPS
as_int

wrong pool in method checkparams again. fixed

^var[^class:var.method[]] is not constructor now

pa_code.h

dox: example1 updated

Methoded reorganized 2. todo: methoded-configure

Methoded reorganized. todo: methoded-configure

removed ^a.menu{$name} ability. now
$a{^menu{$name}}
or
^a.menu{$a.name}

beautifying just compiled. todo: debug, configure

beautifying -99

-d

operators first!
so that ^table.menu{^rem{}} would not be 'unknown column'

for var now written not to r.wcontext, but to r.root
cleared "entered_object" state

libimaging dead end: pil parses header in .py

image:measure -100 just compiled

exec4. todo env

exec -2

table:dir 0 [without regexp]

renamed 'root' to 'op'

operators are now not root methods of parent class. just 'ROOT' class

constructor flag dropped at get_method_frame and remembered into method_frame

sql quote. string untaint UL_SQL

string::pos & cmp & piece bugs fixed.
string::split new
table:load separated from set
table:set implemented with clean \n \t searches

table:join

static|dynamic|any method registration

string: lsplit -1

file:load 1. tainted

fixed r/w context of code-params2, fixed ntfs hardlink slow dir update

fixed r/w context of code-params

about to change junction rcontext!!

sort 0

table:save decided to have as it were. stepped back.
grammar: added 'in' 'is' 'lt'&co follow-space check

disabled $a(123) $b[$a.CLASS] ^b.inc[123] shit.
allowed no checks in native method realisations on 'self' validity -
they now can simply assume that 'self' is V{Proper} .

html-typo 0
html-typo 1. noticed double default content-type prob. run cgi.cmd tomorrow first
html-typo sample. detected << problem: they become ltlt too fast

@CLASS equals @BASE. sanity

cgi: reinvent document_root and request_uri under iis. autoconf updated. introducing ap_config_includes

shifted apache branch on main trunc