Aubit 4GL SVN-Code

configure: restore --with-pg-all, and find PostgreSQL where RHEL puts it

Reported: PostgreSQL is no longer detected from SVN, while the released
tarball detects it and builds. Alma Linux 8, PostgreSQL 18, invoked as

./configure --with-pg-all=/usr/pgsql-18

--with-pg-all was a shorthand in the old configure.in that pointed every
PostgreSQL component at one prefix. The rewrite dropped it, and autoconf
only WARNS about an unrecognised --with-... - so configure completed,
the option did nothing, and PostgreSQL silently went missing. It is
restored here, setting --with-postgresql and --with-ecpg unless those
were named explicitly.

This did not show up on Debian or Ubuntu because libpq.pc is on the
default pkg-config path there, so detection succeeded anyway and masked
the ignored option. On RHEL, Alma and Rocky the PGDG packages install
under /usr/pgsql-NN with nothing on the default PATH or PKG_CONFIG_PATH,
so both pkg-config and a bare pg_config come up empty.

So detection also gained a fallback chain: pkg-config, then pg_config on
PATH, then pg_config under /usr/pgsql-*, /usr/local/pgsql-* and /opt/*
newest first, then a direct search for libpq-fe.h in the usual include
directories. Whatever answers is then verified by compiling against
libpq-fe.h and linking PQconnectdb, so a stale pg_config reports "not
found" rather than breaking the build later.

Verified: --with-pg-all against a /usr/pgsql-18 style prefix gives
-I<prefix>/include and -L<prefix>/lib -lpq; and with pkg-config unable
to see libpq, the pg_config fallback still finds it.

Update the vendored SQLite3 ODBC driver from 2007 to 0.99991

The vendored copy dated from 13 May 2007 (r9422) and segfaulted on any
statement carrying a bind parameter - which is nearly every statement a
4GL program issues - so SELECT ... WHERE id = lv_id or INSERT with
variables crashed while all-literal statements worked.

Two causes, both long since fixed upstream:

It built its SQL by textual substitution, calling
sqlite3_vmprintf(s->query, (char *) params) - an array of char * cast to
a va_list. That is valid on i386, where a va_list is effectively a
pointer to the stack arguments, and invalid on x86-64, where it is a
struct of register save area offsets. GCC 14 rejects the cast outright,
which is why this file was already built with the diagnostic switched
off. Upstream replaced the whole mechanism with sqlite3_bind_*() on
prepared statements years ago.

Its header was also pinned to sqlite3-local.h, which declares SQLite
3.0.8 (2004), while the plug-in links the system libsqlite3 3.45.1.
Anything added since 3.0.8 - sqlite3_prepare_v2, sqlite3_malloc - was
implicitly declared and had its returned pointer truncated on a 64 bit
build.

Upstream's sqlite3odbc.c and .h replace both files, keeping the small
Aubit preamble (the A4GL_debug shim and #define AUBIT). Three build
flags follow upstream's renames: WITHOUT_DRIVERMGR became
USE_DLOPEN_FOR_GPPS, WITHOUT_WINTERFACE avoids needing unixodbc-dev, and
the in-tree ODBC header gains SQL_C_WCHAR, a standard constant it
predates.

Parameterised statements, cursors and quote escaping now all work.

Generator: unblock report modules and fix generated label names

Measured against fgldemo - 41 modules, 5900 lines - rather than guessed
at. Two fixes, both found by trying to compile it:

binding.o was never in the plug-in's link line, and binding.c is in any
case wrapped entirely in #ifdef CRAP, so A4GL_dtype_sz was never
compiled even though compile_cs_sql.c calls it. Every REPORT module
failed at dlopen with an undefined symbol. The guard now closes before
the dtparts table, leaving decode_datetime and A4GL_dtype_sz live - the
only two the caller needs - and the two SPRINTF macros in them, whose
header is also inside the disabled part, are plain snprintf now.

Generated labels used %d on a value that can be negative, producing
continue_input_-908174256, which is not a C# identifier. Now %u.

fgldemo now generates 39 of 41 modules, up from 36. The two that do not
are LOAD/UNLOAD, which are assertion stubs in the generator, and
DROP DATABASE, which Aubit refuses everywhere and is not a C# issue.

.NET: 4GL's implicit conversions, display widths, and a README

4GL allows LET lv_char = lv_int and converts on the way; C# does not. The
generator already resolves this at compile time, emitting an explicit
conversion only where the declared types differ - LET b = a between two
INTEGERs still emits a plain assignment. This adds the conversions it
emits: AsChar, AsString, AsInt, AsSmallInt, AsBigInt, AsDecimal, AsFloat,
AsSmallFloat, AsBool, AsDate, AsDateTime, AsMonthSpan and AsTimeSpan.

They are hand-written rather than System.Convert, which is wrong in the
three places that matter and has a test spelling it out: Convert rounds
7.9 to 8 where 4GL truncates to 7, turns NULL into 0 where 4GL keeps
NULL, and throws on "abc" where 4GL yields NULL.

DISPLAY now passes each value's width, not just a DECIMAL's precision.
4GL prints an INTEGER in 11 columns, a SMALLINT in 6, a FLOAT in 14 - and
a NULL still occupies its field, which a boxed null cannot know on its
own. CHAR is deliberately excluded: it is already padded at assignment,
and widening it again would defeat CLIPPED.

A fifth end-to-end program, convert.4gl, covers thirteen cross-type
assignments including truncation of negatives and NULL propagation. All
four end-to-end programs and the other four probes pass.

Adds dotnet/README.md.

Generator: carry DECIMAL precision and scale; control flow probe

Same problem as CHAR, same shape of fix. A DECIMAL(10,2) maps to C#'s
decimal, which carries neither the declared precision nor the scale, so
2.5 printed where 4GL prints 2.50, and it printed bare where 4GL right
aligns it in 12 columns.

Both are supplied where the declaration is still known. The assignment
wrapper - already used for CHAR - now also emits

L_d = Fgl.Scale(expr, 10, 2);

which rounds to the declared scale and forces the trailing zeros, since
Math.Round reduces a scale but never extends one. The width is the
declared PRECISION plus two and is passed at the DISPLAY site, where
expr_datatype still knows it:

Display("div =", Fgl.Disp(L_d, 10));

Adds IsNull, AsDecimal and CompareString to the runtime, which generated
code calls unqualified.

A third end-to-end probe covers FOR, WHILE, IF, CASE, arithmetic, string
concatenation, a function with a return value, a compound condition and
IS NULL. All three end-to-end probes pass, as do the other four and the
188 unit tests.

Generator: carry CHAR widths through to the generated code

A 4GL CHAR is fixed width and blank padded, but it maps to a C# string,
which has no width - so LET c = "hello" on a CHAR(10) was losing the
padding and DISPLAY printed five columns instead of ten.

The width is known at the assignment even though it is not recoverable
from the C# type, so every assignment to a CHAR is now wrapped:

L_s = Fgl.Pad("hello", 10);

This uses the generator's existing write-prefix / suffix mechanism in
decode_varbind, so it costs one helper and two small branches rather
than switching the emitted type to FglChar - which would have touched
the declaration, LET and parameter paths.

VARCHAR is deliberately excluded: it has no fixed width to pad to.

Also adds Fgl.CompareString, which generated code calls wherever either
side of a comparison is a string. Trailing blanks are not significant,
and a comparison involving NULL is false - including NULL against NULL.

A second end-to-end probe covers nine assignment forms: CHAR padding and
truncation, CHAR from CHAR, VARCHAR, an array element, a record field,
NULL, and an INTEGER's display width. Both probes pass.

Add the C# generator plug-in source (lex_cs)

lex_cs was copied into the tree but never added to SVN, so it - and the
retargeting in r13050 - existed only in a working copy. This adds the 16
source files the plug-in builds from, 432K in total.

Deliberately NOT added: the directory also holds about 35M of backup
copies (latest/, 250210/, CM/, old1/, backup_221009/), seven copies of
the same 1.7M tarball, and .exe/.dll/.mdb binaries including an Informix
DLL. Those are ignored rather than versioned; they are still on disk if
any of them turn out to matter.

The plug-in is still not built by the top-level make - it has to be
built by hand in lib/liblex/lex_cs. Wiring it in is a separate decision,
since it would make every build depend on it.

Retarget the C# generator onto the .NET runtime

The CSNEW generator already built and worked - it was simply never wired
into the top-level build. Retargeting it turned out to be a handful of
edits rather than a rewrite:

namespace and using -> Aubit4GL / Aubit4GL.Generated
base class -> FglModule, which carries the session
DATE -> DateOnly, an exact match
DECIMAL and MONEY -> decimal (this is where unknown_8 dies)
DATETIME, INTERVAL -> FglDateTime, FglInterval
BYTE and TEXT -> byte[]
SMALLINT -> short, not int
CLIPPED, LENGTH,
UPSHIFT and friends -> Fgl.*

DISPLAY now emits its expressions as separate arguments rather than
concatenating them, because 4GL renders each value at its type's own
width - an INTEGER in 11 columns, a SMALLINT in 6 - and joining them
into one string first threw that away. SMALLINT mapping to int? had the
same effect and is fixed.

A new end-to-end probe compiles ONE 4GL program through BOTH toolchains
and diffs what they print. It is the only check that exercises the
generator and the runtime together; every other probe tests a
hand-written .NET side. It passes.

Known gap, recorded at the head of the probe: CHAR(n) maps to a bare
string, so its declared width is lost and an unclipped CHAR prints
without its padding. Fixing that means carrying the width into the
declaration and routing assignment through Assign(), which touches the
declaration, LET and parameter paths.

.NET: REPORT engine

Page and line tracking, the four margins, COLUMN positioning, SKIP, NEED,
page breaks, and the aggregates. A probe compares a report's output
against the C runtime line for line - nothing normalised, since a report
is entirely about column and page layout - and all 66 lines match.

The state follows BaseFGLReport from the original C# library, which had
the data model right even though the behaviour was never written. The
layout defaults were read out of the generated C rather than guessed:
page length 66, margins top 3, bottom 3, left 5, right 132.

COUNT(*) and SUM/AVG keep separate counters: COUNT counts rows while the
others only see non-null values, so a row with a NULL still counts but
does not drag the average down.

.NET: USING format engine

Thirty formats now render identically to the C runtime, checked by a new
probe. Padding is significant here, so values are bracketed and compared
exactly rather than trimmed.

Four rules the probe corrected me on, all of which look wrong coming
from another language:

0 is a LITERAL, not a digit placeholder - the zero-fill character is &.
"###,##0.00" is valid in Oracle and means something else entirely in
4GL, and the C is right to render 1234.5 as 12,340.00 for it.

A zero integer part is blank under #: 0 with "###" is three spaces.

Fill is per position, not per field - one & in "##&" does not zero-fill
the other two.

Sign and currency characters occupy digit positions, so "-----" is five
of them; treating them as decoration overflows the field to asterisks.
Parentheses are the exception and never hold a digit. A single sign sits
where the mask puts it, a run of them floats to the first digit.