You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 2 Next »

Major BBS v10

Applies to: v10

Editor's note: this chapter is new in the v10 layer. Chapters 1-28 are the drafts/wg3 chapters with v10 notes inserted; this chapter collects what a module developer must know that does not belong to one place. It is not Marc Fountain's text. Section 29.1 reproduces MBBS10.txt ("The Major BBS v10 Development Guide", version 0.0.1, 14 February 2026) in full. Everything after it is checked against the v10 headers in inc/ (branch add-dntapi2-header, HEAD 1c02973; the working tree also carries an uncommitted #include "dntapi2.h" in GCOMM.H and two forward declarations in LOCKNKEY.H). Citations are FILE.H:line in that tree. Anything the headers cannot prove is marked unverified.

29.1 MBBS10.txt (verbatim)

Applies to: v10

1. Overview. This document needs to grow into a holistic Confluence wiki page and/or document file that describes the full Major BBS v10 API along with the GSBL routines. For now, it's just a guide to using the historical documentation - what to leverage and what to ignore (or use knowing there are changes).

2. Worldgroup 3.2 Similarities. Major BBS v10 is the next version from Worldgroup 3.2 (which in sequence would have been Major BBS 9.2). As such, the API is nearly identical. There are a few changes to be aware of:

  • The "Client/Server" (Visual Basic app) mode has been removed and is no longer supported.
  • The Active HTML mode is no longer supported, mainly because the C++ code has yet to be ported. It is possible it will return, or it will be replaced with a better React or HTMLX front end capability with Websockets.
  • The rtihdlr() routine has been restored and is functional. This is the realtime interrupt handler that used to be limited to DOS versions of the BBS. It's been returned to the core system with native Windows routines and old modules using it can now be ported!

3. Differences. The compilation IDE is Visual Studio, currently using VS2022 at the time of this writing. This has some serious impacts that you need to keep aware of. First, the language reference moves from C89 to C23 (ISO/IEC 9899:2024). Worldgroup 3.2 was compiled with Borland C/C++ 5.0. Borland added a number of things that made it to C99, but was not necessarily standard. Most important is that VS2022/C23 are far more strict, and loose memory management will fail spectacularly.

One particular issue is string literals. While Borland would let this pass, VS2022 will not. We have macros in our code base that help with this -- but you can modify it how you prefer. String literals are "const char *" and as such cannot be modified. Passing them and using them like "char *" is not a good idea and will crash the system. The GSBL routine btuxmt(channel, string) was often used with string literals, or literals from a #define. You need to pass them as true char *. spr() works to take a const char * and output a char *, because it is the Galacticomm version of sprintf(). A str2buf() routine would work much in the same way (without the ability to parse a variadic template string).

There is a Borland<->VS2022 DLL that allows Borland-compiled modules for Worldgroup 3.x for NT to work. Usually it works just fine, but sometimes a module needs to be converted to native Version 10.0.

Finally beware structure packing. You should be familiar with this if you ported DOS modules from Turbo C 2.0 or Borland C 3 or 4, but if not, beware the size of your structs and the offsets of their members.

If you'd like to debug your add-on in Release mode, have the bbs running, and in VS2022 select Debug and then attach to process, select wgserver.exe. Unfortunately there are bugs that crop up only in Release (Debug will zero out pointers so garbage memory does not always turn up - same with string literal issues).

Applies to: v10

Header check: str2buf() is not declared anywhere in v10 inc/; the text proposes writing one ("would work much in the same way"). It is not an API. The project's own helper is CHAR_BUFFER(name, size, lit), which expands to CHAR name[size] = lit (GCOMM.H:40).

29.2 What was removed, what stays

Applies to: v10

Subject Status in v10 Evidence
Client/Server mode (Visual Basic client apps, agents, Worldgroup Manager) Not in v10 MBBS10.txt section 2; tlc.h:112 ("AHTML and C/S Support removed", under #ifdef SUPPORTREMOVED); the v10 repository history has the commit "Initial port of GALTELE including removal of AHTML and CS related code".
Active HTML Not in v10 MBBS10.txt section 2. The v10 headers confirm it: the TLCAH.CPP declarations (iAHTrans, initactiveh, initActiveHLite, initahpag) sit inside #ifdef SUPPORTREMOVED //AHTML and C/S Support removed (tlc.h:112-120), so they are compiled out; likewise the AHTML hooks tlcenter()/tlcexit() (TLCAPI.H:69-70). GALACTH.H is still shipped, so the presence of that header says nothing about support. SUPPORTREMOVED is the v10 port's marker for removed AHTML and C/S code (also tlcact.h:21, tlcchan.h:19, tlcpal.h:19).
WGM stand-alone client tooling unverified MBBS10.txt names only "Client/Server (Visual Basic app) mode". The v10 vcproj/ has a project named wgsmkwgm; its status is not established. Treated here as client side, hence not in v10.
rtihdlr() present; "restored and functional" per MBBS10.txt WGSEXP VOID rtihdlr(VOID (*rouptr)(VOID)); (MAJORBBS.H:746). The prototype is identical in WG3.2 (MAJORBBS.H:750 there), so the claim is about run-time behaviour: unverified from headers. The tick rate (the "18 Hz" of chapter 11) is unverified.
GSBL (galgsbl) present EXPGSBL macros (BRKTHU.H:20-28). The DOS hardware parts of chapters 13-19 are historical.
X.25 no X.25 header in v10 inc/ (listing checked) chapter 19 is historical; unverified whether any X.25 support remains.

29.3 Building a v10 module

Applies to: v10

Everything in chapters 1, 2, 21 and 23 about Borland C++, Phar Lap, .MAK files, MKNT, \WGDEV\BIN and InstallShield describes the WG1-3.x environment. For v10 (sources: MBBS10.txt section 3, analysis/v10/build-notes.md):

  • IDE and project files. Visual Studio; MBBS10.txt says VS2022, the repository's projects name PlatformToolset v145 (VS 2026). Every one of the 138 .vcxproj builds both Win32 and x64.
  • Language. MBBS10.txt says C23 (ISO/IEC 9899:2024), but no project file selects a C standard (no LanguageStandard_C in any of the 138 projects; ConformanceMode is true), so what a module actually gets is the MSVC default C mode for .c and C++ for .cpp: unverified against the C23 claim.
  • Plain char is unsigned. 118 of the 138 projects pass /J, preserving the Borland default. Do not assume a signed char.
  • Structure packing is chosen by the project, not the header: StructMemberAlignment 8Bytes on 86 projects, and no v10 header uses #pragma pack (the only pragmas are #pragma inline, CRIT.H:33, and #pragma message, FUSION.H:30). A struct copied between a DOS-era module and a v10 module will not necessarily have the same size or offsets. Never put pointers, HANDLE, size_t or unsigned __int64 into a Btrieve/DFA record: use the fixed-width engine types (GBOOL is short, GCTYPDEF.H:135; SHORT/USHORT are short; LONG/ULONG are long, 32-bit on Windows x64 as well, GCTYPDEF.H:108-109).
  • Typical module defines (from the shipped elwbio project): WIN32;GCWINNT;GCMVC;BBSVER=1000;USE_DEF_FILE;_CRT_SECURE_NO_WARNINGS;...;__BUILDV10MODULE. Include inc and link wgserver_lib.lib (plus galgsbl_lib.lib for GSBL users). BBSVER is a leftover: the header constant is now BBSVERS (server.h:61).
  • Exports. With USE_DEF_FILE defined EXPORT, IMPORT and EXPORTV are empty and the module must export through a .DEF file (GCTYPDEF.H:60-64); without it they are __declspec(dllexport)/dllimport (GCTYPDEF.H:66-69). A module entry point written void EXPORT init__xxx(void) (chapters 2, 3, 5, 6) therefore compiles either way, but a module built with USE_DEF_FILE needs its init__xxx in the .DEF.
  • Deploy and debug loop (repository README.md, not MBBS10.txt): install a working v10, start the BBS once and stop it; run Visual Studio as Administrator with startup project wgserver, command wgserver.exe, arguments /regserver; build and copy the output DLLs over the installed ones; a fresh instance reporting Btrieve Error #20 needs WGSAPPGO from an elevated prompt.
  • Debugging a Release build: start the BBS, then Debug -> Attach to Process -> wgserver.exe (MBBS10.txt). The statement that Debug zero-fills pointers is the author's observation: unverified.
  • Borland modules. MBBS10.txt says a Borland<->VS2022 DLL lets Borland-built WG 3.x/NT modules run. The tree has a forwarding CW3220MT DLL (replacement for the Borland runtime CW3220MT.DLL, forwarding to CW3220BCC), which fits the description, but MBBS10.txt does not name it: inference.
  • Heap ownership. alcmem, alczer and alcrsz are the engine's allocator (alczer, GCOMM.H:287; getml(ln) maps to alcmem(ln), GCOMM.H:94). Do not free() their results with the C runtime, and do not hand C-runtime memory to engine routines that free it. The rule is general; the engine source is not in the tree: unverified.
  • Release-only crashes. MBBS10.txt: "Release ... bugs that crop up only in Release (Debug will zero out pointers so garbage memory does not always turn up - same with string literal issues)". See 29.4.

29.4 HAZARD: const char * string literals versus char * parameters

Applies to: v10

Read this before porting any code. MBBS10.txt: "String literals are 'const char *' and as such cannot be modified. Passing them and using them like 'char *' is not a good idea and will crash the system. The GSBL routine btuxmt(channel, string) was often used with string literals, or literals from a #define. You need to pass them as true char *."

Most of the printed examples in this guide (chapters 3-6, 11, 16, 17) pass literals, or #defined literals, to routines declared with a plain CHAR *. The v10 headers still declare those routines without const:

EXPGSBL INT btuxmt(INT chan,CHAR *datstg);     /* BRKTHU.H:162 */

The same header declares the sibling btucmd correctly: EXPGSBL INT btucmd(INT chan,const CHAR *cmdstg); (BRKTHU.H:159). So btucmd(chan,"ATZ") is safe and btuxmt(chan,"ATDT13055837808\r") is the call that MBBS10 warns about (chapter 16 prints exactly this form).

Why it fails

  • In C, a string literal has type char[N] (not const), so passing it to a CHAR * parameter compiles without a diagnostic on MSVC too. The failure is at run time, when the callee writes to it: the literal lives in a read-only string pool, and the write is an access violation. This is why such code builds cleanly and then crashes, typically in Release (MBBS10.txt).
  • In C++ (.cpp projects, ConformanceMode true) the same call is a compile error.
  • These two points are general MSVC behaviour, not statements from the sources; the sources state only the crash.

The project's escape hatches

Helper Prototype Where Use
CHAR_BUFFER #define CHAR_BUFFER(name, size, lit) CHAR name[size] = lit GCOMM.H:40 a writable local copy of a literal
spr CHAR *spr(const CHAR *ctlstg, ...) GCOMM.H:404 Galacticomm sprintf(); takes a const format, returns a writable CHAR * (a shared internal buffer: copy it if you need it after the next call; that convention is unverified in v10 source)
stzcpy CHAR *stzcpy(CHAR *dst, const CHAR *src, UINT num) GCOMM.H:409 copy with limit and zero fill
gstrcpy gstrcpy(CHAR *destination, const CHAR *source) GCOMM.H:1211 wrapper for the C library copy

A. Mutable CHAR * parameters that are probably only read

The compiler lets a literal through; whether each callee writes is unverified (no engine source in the tree). Until shown otherwise, pass a local CHAR buf[N] or a CHAR_BUFFER.

v10 prototype Where
EXPGSBL INT btuxmt(INT chan,CHAR *datstg); BRKTHU.H:162
EXPGSBL INT btuxmn(INT chan,CHAR *datstg); BRKTHU.H:210
WGSEXP VOID snd2ch(INT chan,CHAR *text); WGSEXP VOID snd2al(CHAR *text); MAJORBBS.H:1106-1107
WGSEXP VOID addanno(CHAR *message); WGSEXP VOID addannom(CHAR *message,CHAR *appid,CHAR *cmdstg); MAJORBBS.H:867, 866
WGSEXP VOID bootem(CHAR *who); WGSEXP INT entmdl(INT modnum,CHAR *cmdstg); WGSEXP VOID shochn(CHAR *legend); MAJORBBS.H:863, 910, 1086
INT stricmp(CHAR *,CHAR *); INT strcmpi(CHAR *,CHAR *); INT gstricmp(CHAR *,CHAR *); INT memicmp(CHAR *,CHAR *,size_t); GCOMM.H:70, 71, 69, 73 (they look like the C library functions but are not const-qualified)
findstg, wcmatch, setcnf, getoption, strsrep (CHAR * parameters) GCOMM.H:249, 1080, 555, 747, 433
dspmsg, dspmlt (CHAR *parm1..parm3) BBSUTILS.H:56, 63

B. In-place modifiers: a literal here is a write to read-only memory

The header comment says the string is altered.

v10 prototype Header comment Where
CHAR *strupr(CHAR *); CHAR *strlwr(CHAR *); "make string uppercase" GCOMM.H:67-68
fixpth(CHAR *path) "add a backslash to string if needed" (the buffer needs room for it) FIOAPI.H:300
stpans(CHAR *str) "strip ANSI sequences from a string" GCOMM.H:573
strstp(CHAR *str,CHAR ch) "strip a character from a string" GCOMM.H:582
uniqfn(CHAR *fspec) "create a unique filespec" FIOAPI.H:315
xltctls(CHAR *txtbuf) translate control characters in a buffer MAJORBBS.H:747

C. Correctly const-qualified (safe with literals)

prf(const char *fmt, ...) (GCOMM.H:823); shocst(const CHAR *brief, const CHAR *detail, ...) (MAJORBBS.H:1080); btucmd(INT, const CHAR *) (BRKTHU.H:159); opnmsg(const char *mcvfil) (MCVAPI.H:54); dfaOpen(const CHAR *filnam, USHORT, const CHAR *owner) (DFAAPI.H:217-220); makePath and the isdir/isfile/fndfile family (FIOAPI.H); stzcpy and gstrcpy (above). The exact text of every prototype named in this section was extracted from inc/ by tools/v10_crosscheck.py and is listed in CROSSCHECK.md ("Hazard prototypes, as printed in the v10 headers"); where this table and that listing disagree, the listing wins.

D. Rules

  1. Never pass a literal, or a #define that expands to one, to anything in table B. Copy it into a CHAR_BUFFER or array first.
  2. For table A, use a local buffer until the callee is shown not to write. Do not const-qualify your own wrapper unless you can prove the callee never writes.
  3. When you need to format, use spr() rather than passing the literal.
  4. Never cast away const ((char *)"abc") to silence a warning: that hides exactly this bug.
  5. Compile your module once as C++, or raise the warning level, in Debug to surface the call sites.
  6. A #define is a literal. #define PROMPT "Enter: " followed by btuxmt(chan,PROMPT) is the case MBBS10.txt names.

E. Related v10 prototype hazards

  • stpcpy: the comment says "copy a string with limit" but the prototype CHAR* stpcpy(CHAR* dst, CHAR const* src) has no limit parameter (GCOMM.H:576-579). No bounds checking.
  • MAXPATH is _MAX_DRIVE+_MAX_DIR+_MAX_FNAME+_MAX_EXT with no parentheses (FIOAPI.H:96, _MSC_VER only): parenthesize before using it in arithmetic.
  • char* results of getCPUInfo(), getOSVersion() and formatBytes() (OPRLOW.H:152-155): who owns the buffer is unverified; do not write into or free them.
  • findfirst/findnext and unsigned __int64 returns exist only for _MSC_VER (FIOAPI.H:257-269, OPRLOW.H:148-151).

29.5 The export macros changed (breaking)

Applies to: v10

All in GCTYPDEF.H. Module code that only writes EXPORT (as every example in this guide does) is unaffected; code that declares its own exported variables or functions through the macros is not.

WG3.2 v10
EXPORT_VARIABLE(inout,t), EXPORT_FUNCTION(inout,t,n) EXPORT_VARIABLE(t) is extern t EXPORT (GCTYPDEF.H:161); EXPORT_FUNCTION(t, n) is extern t (* EXPORT n) (:162); MSVC variants EXPORTV t / EXPORTV t (* n) (:167-170); other platforms extern t (:178-181)
IMPORT_VARIABLE(t), IMPORT_FUNCTION(t,n) built on EXPORT_*(IMPORT_TYPE,...) IMPORT_VARIABLE(t) is extern t IMPORT (:163); IMPORT_FUNCTION(t, n) is extern t (* IMPORT n) (:164)
EXPORT_TYPE / IMPORT_TYPE removed; use EXPORT, IMPORT, EXPORTV, IMPORTV (:60-69)
(none) DLLAPI = __stdcall, CPPAPI = __cdecl (:56-57)

Other consequences: each per-module export macro (GMEEXP, DNSEXP, FILEXP, TCPIPEXP, ...) is now EXPORT/IMPORT instead of __export/__import, and struct tags carry it (struct GMEEXP qscfg {, GME.H:73); EXPGSBL is EXPORT or IMPORT depending on __BUILDGSBL (BRKTHU.H:20-28); the names EXPWGSF, EXPWGSV, EXPRSYF, EXPRSYV and TLCEXP (now TELEEXP, TLCAPI.H:28) are no longer defined (TLCEXP still appears, undefined, at tlc.h:115 inside the compiled-out SUPPORTREMOVED block); new paired macros XXXEXPF(x,y)/XXXEXPV(x) exist, for example WGSEXPF/WGSEXPV (GCOMM.H:22-27). __export/__import survive only under __BORLANDC__ (GCTYPDEF.H:74-77).

29.6 Sizes, layouts and constants

Applies to: v10

Item WG3.2 v10 Evidence
struct ffblk.ff_name char[256] char[260] FIOAPI.H:44
struct xffblk absent { struct ffblk *fbptr; HANDLE ff_handle; UINT s_attrib; WIN32_FIND_DATA nfblock; } FIOAPI.H:50-55. Holds a HANDLE: its size differs between Win32 and x64
MAXPATH, DRIVE, DIRECTORY, FILENAME, EXTENSION not in FIOAPI.H _MSC_VER only FIOAPI.H:96-100
GCSTRPTH, GCMAXPTH MAXPATH, or 261 under GCV2/GCDOS FIOAPI.H:103-111
MAXPARSZ MAXSFORS under STARTER, else (MAXBLK/sizeof(VOID *)) only (MAXBLK/sizeof(VOID *)) GME.H:61; MAXBLK is 65530U (GME.H:57) and sizeof(VOID *) is 8 on x64, so the element count is 8191 (arithmetic from the macros, not run)
GMEWRKSZ 518 under GCDOS, else 1536 same GME.H:42-44
TPCSIZ GME.H 51 (the teleconference header ENTTLC.H had 41) 51 GME.H:48; not a v10 change (29.8)
DFAFILE struct dfablk struct dfablk DFAAPI.H:167; unchanged
BBSVER / BBSVERS BBSVER 330, "3.30" BBSVERS 1034, BBSVERSSTG "10.34" in the Major BBS branch of server.h server.h:61-62; renamed (extra S). BBSVER is still defined, with other values, in the WEBCAST (200) and ACTIBASE (100) product branches (server.h:31, 46), which a Major BBS module does not select
MAXSUSER, MAXSUSR, MAXCALLS, GENMODL, SSYSUID MAJORBBS.H:27-30 (3.2) absent starter-edition limits are gone
DMAUSR absent 0x08000000L, "User has logged in via GALDMA" MAJORBBS.H:282

29.7 Headers dropped and added

Applies to: v10

118 headers are common to WG3.2 and inc/ of v10. Not in the shared v10 inc/ (30 files of the WG3.2 INC): 2030.h, ch.h, enttlc.h, etlact.h, etlapi.h, etlcs.h, etlfsd.h, etlfuncs.h, flash.h, galbj.h, galdroid.h, galentl.h, galetl.h, galfaz.h, galhang.h, galkyrl.h, galkyrm.h, galkyrs.h, gallst.h, galpok.h, galsnova.h, galtng.h, gamutils.h, gcsasys.h, kyrandia.h, kyrspls.h, layer.h, mr_mail_vir.h, wgsmajor.h, wortlc.h. "Not in inc/" is not "removed from the repository": searching the whole v10 tree finds wgsmajor.h/wortlc.h under inc/msg/wg32/ (message-number headers, so #include "wgsmajor.h" needs that directory on the include path, which the module projects have: $(SolutionDir)inc\msg\wg32), project-local copies of ENTTLC.H, ETL*.H, FLASH.H, galetl.h in src/isv/galetl/, of GCSASYS.H in src/apps/galrsy and src/server/wgserver, of LAYER.H in src/isv/mjwmut and src/isv/mjwwhl, and of the Kyrandia/GAMUTILS.H headers in src/isv/elwkyr/elwalc2. A new module that includes only the shared inc/ cannot use any of them. Consequences for this guide:

  • ENTTLC.H and the ETL*/GALETL/WORTLC.H stack, the old C teleconference: not shared. The v10 teleconference is the C++ tlc*.h set. Code coded to ENTTLC.H needs a port (or the local copy in src/isv/galetl/).
  • GCSASYS.H (hasaxs, copyuser): not shared, and with it the ACCESS(op) macro.
  • 2030.H is nowhere in the v10 tree. LAYER.H, the Btrieve-name compatibility layer that chapters 9 and 28 tell you to include, is not in the shared inc/; the ported modules carry their own local copy (src/isv/mjwmut/LAYER.H, whose text differs from the WG3.2 file). Copy it into your project if you want the *btv macros, or use dfa*() from DFAAPI.H directly. Whether GALPORT-style renames compile is unverified.
  • The game SDKs (KYRANDIA.H, GALKYR*.H, GALSNOVA.H, GALPOK.H, ...) are not engine headers.

Added to inc/ (6): DMA.H (GALDMA direct module access server), DNTAPI2.H, PIRACY.H (internal, deliberately not documented), SCRIPT.H, SMNNHOOK.H, WGSMSGS.H (ELOG_* event-log ids).

Removed from the common headers: only worAliasOnline (WORCOMN.H:339 in WG3.2; absent from v10).

29.8 Things that look like v10 changes but are not

Applies to: v10

The automatic delta (analysis/deltas/SUMMARY.md) over-reports. Do not "fix" your code for these:

Claim What the headers say
C23 removed default arguments in palSave, palStatus, tlcChannel*, tlcTrans*, tlcPublicSendAllChannels False. v10 still has them: VOID palSave(INT unum, GBOOL updateself = TRUE); (tlcpal.h:59), const CHAR* user = NULL (tlcpal.h:63), TLCAPIC.H:62-65, 177-178. The two trees differ only in spacing around =. These are C++ headers: a C build cannot include them, with or without C23.
TPCSIZ changed 41 to 51 GME.H is 51 in both trees (GME.H:48). The 41 is the teleconference header ENTTLC.H:32 (#undef TPCSIZ / #define TPCSIZ 41), which v10 does not ship. The 40 printed in chapter 5's sample program is that sample's own define.
tlc flags renumbered (XTOOTH 128 to 0x0001, TYPING 512 to 0x0004, RSTX2M 1024 to 0x0002) tlc.h:48-50 has 0x0001, 0x0002, 0x0004 in v10, and WG3.2's tlc.h already had the same values. Only code written against the dropped ENTTLC.H (128/512/1024) sees a difference.
DFAFILE changed from BTVFILE to struct dfablk DFAAPI.H:167 is #define DFAFILE struct dfablk in both. BTVFILE existed only in the compatibility headers (2030.H, LAYER.H), which are not in the shared v10 inc/ (29.7).
MAXPARSZ redefined non-STARTER value unchanged; the STARTER variant is gone (29.6).
"34 structs removed" 16 belong to dropped headers; the other 18 (qscfg, dns, dnshstex, pageInfo, ...) are still defined, now with the module export macro in the tag. Parser artifact.

The one real struct change is struct ffblk (29.6).

29.9 New APIs

Applies to: v10

API Prototype / content Where
HookAdd GBOOL HookAdd(USHORT hooktype, voidfunc hookfunc, SHORT priority); priority runs 32767 to -32768 SMNNHOOK.H:30-34
HookNext INT HookNext(USHORT hooktype, INT curidx); returns an index or NOIDX; pass NOIDX for the first SMNNHOOK.H:36-39
struct HookItem, HookArr, nHooks SMNNHOOK.H:20-27
getCPUInfo, getOSVersion, formatBytes return char *; buffer ownership unverified OPRLOW.H:152-155
getTotalDiskSpace, getFreeDiskSpace, getTotalMemory, getFreeMemory unsigned __int64 OPRLOW.H:148-151
IsWow64, Is64BitWindows BOOL OPRLOW.H:147, 154
findfirst, findnext Borland-style over Win32 FindFirstFile, _MSC_VER only; fnd1st/fndnxt remain FIOAPI.H:257-269
gstrcpy gstrcpy(CHAR *destination, const CHAR *source) GCOMM.H:1211
script engine begscr, endscr, abtscr, pumpscr, go2lbl, psnstg, screrr, scrinp, register_scrcmd, init__galscr; constants NUMWAT 8, MAXELN 21 SCRIPT.H
GALDMA direct module access server; RLOGINSTART 1000; user flag DMAUSR DMA.H:42, MAJORBBS.H:282
Modern date/time DNTAPI2.H: dnt2_* functions, prndat2, prnday2, prntim2 present on branch add-dntapi2-header only, not in main; chapter 27 note
CHAR_BUFFER writable local copy of a literal GCOMM.H:40
WGSMSGS.H ELOG_* event-log message ids WGSMSGS.H

29.10 Checklist for porting a WG3.2 module to v10

Applies to: v10

  1. Replace BBSVER with BBSVERS (BBSVERSSTG for the string) when you target the Major BBS branch of server.h.
  2. Replace uses of EXPORT_TYPE, IMPORT_TYPE, EXPWGSF/V, EXPRSYF/V, TLCEXP and the two-argument EXPORT_FUNCTION/EXPORT_VARIABLE (29.5).
  3. Re-derive anything sized from ff_name[256] (29.6).
  4. Remove includes of the headers that are not in the shared inc/ (29.7), or carry a project-local copy as the ported modules do.
  5. Audit every call that passes a string literal or #define to a CHAR * parameter (29.4).
  6. Decide your struct packing and check on-disk record sizes (29.3).
  7. Do not remove default arguments from tlc*/pal* calls (29.8).
  8. Drop all Client/Server and Active HTML code (chapters 20, 21, GALACTH.H users).
  • No labels