• src/sbbs3/userdat.c

    From Rob Swindell (on Debian Linux)@1:103/705 to Git commit to main/sbbs/master on Friday, October 02, 2026 19:25:26
    https://gitlab.synchro.net/main/sbbs/-/commit/d83fbe2a41580e37314ea671
    Modified Files:
    src/sbbs3/userdat.c
    Log Message:
    Reject future birthdates in birthdate_is_valid()

    The validator had a lower bound (1900) but no upper bound, so a birthdate
    such as 09/11/2890 was accepted at new-user registration and stored,
    and getage() then reported a negative age (-864).

    A birthdate later than today is now invalid. Every entry path that
    validates (new-user and profile prompts via system.check_birthdate())
    rejects it, and getage() returns 0 for any such record already stored.

    Fixes #1277

    Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
    --- SBBSecho 3.38-Linux
    * Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)
  • From Rob Swindell (on Debian Linux)@1:103/705 to Git commit to main/sbbs/master on Monday, November 09, 2026 17:12:39
    https://gitlab.synchro.net/main/sbbs/-/commit/b643ac2bca589181742b3e75
    Modified Files:
    src/sbbs3/userdat.c
    Log Message:
    logoutuserdat(): unlock the user record before closing the file

    Since 48922a15c6 (2025-12-17), logoutuserdat() has read the user record
    with leave_locked=true, written it with fputuserdat(), and closed the file without unlocking the record, relying on close() to release the lock. Every other user record writer unlocks explicitly.

    On Windows, the release of a byte-range lock still held at close happens whenever the OS gets to it (Microsoft's LockFileEx documentation recommends
    an explicit unlock). The web server calls logoutuserdat() after nearly every authenticated request, so a webv4 user firing parallel requests could leave their record locked long enough for the other request threads to exhaust
    their lock retries and log "!Error -104 writing user data for user #N".
    On Vertrauen's Windows web server, every request from one such user stalled
    for about two minutes before failing this way.

    That commit also dropped the dirtyuserdat() call that the previous putuserstr()/adjustuserval() writes had performed; restore it.

    Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
    --- SBBSecho 3.38-Linux
    * Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)