Windows: official port -- UCRT static build, UTF-8 throughout, Low IL sandbox, ncurses TUI, installer; v0.6.0

Build * cross build switches to pkgsCross.ucrt64 (shell-win.nix); curl (Schannel), winpthreads and ncurses 6.6 are linked statically, so the release is cst-agent.exe + busybox.exe (busybox-w32 Unicode) with no extra DLLs * windres embeds a manifest with activeCodePage=UTF-8 and a version resource * ncurses 6.6 Windows console driver patches (mouse button state, A_DIM, wide-char declarations): scripts/patches/ * scripts/bootstrap_msys2.sh builds the same release natively in MSYS2 UCRT64; PR #1's MINGW64 notes now point there * vendor/tre-regex (musl's TRE) provides <regex.h> on MinGW UTF-8 * paths cross the system call boundary as UTF-16 (*W APIs); console code pages set to 65001 and restored on exit; files are read and written in binary mode * child output that is not valid UTF-8 is decoded from the console, OEM or ANSI system code page * read_file hides CR; edit_file, apply_patch and write_file keep CRLF line endings and a UTF-8 BOM OS layer and processes * os_dir_* and cst_fnmatch replace the POSIX-only directory walks in subagent, plugins, skills, undo and fs_tools; realpath / path_under and symlink rejection work on Windows * every child goes through CreateProcessW with its own environment block inside a Job Object, so timeouts and Esc end the whole process tree Sandbox (auto mode, "auto · lowil") * run_shell runs with a restricted Low integrity token; the workspace gets an inheritable Low label once, the sandbox TEMP is labeled Low, and the credentials file gets a no-read-up label * drive roots, the profile and system folders are never labeled * network commands still ask, since Low IL does not block the network; Access denied and MSYS2's 0xc0000142 offer a rerun outside the sandbox * /sandbox lists labeled folders, /sandbox reset-labels restores them TUI * raster console fonts are switched to Consolas / NSimSun while running * the home logo falls back to text when the console font lacks U+2580 * mouse drag copy goes to the Windows clipboard Distribution * scripts/install_windows.ps1 installs a Gitea release into %LOCALAPPDATA%\Programs\cst-agent and adds it to the user PATH; cst-agent-update.cmd reruns it; scripts/release_windows.sh uploads the zip with the token taken from the environment * scripts/run_windows.bat is gone; docs/README-windows.md is rewritten Tests * new test_lowil, test_win_utf8 and test_os_dir; tests made portable (cst_test_util.h) * scripts/win_test.sh and scripts/winvm/ run the suite on English and Chinese Windows Server VMs as Administrator, a standard user and a Chinese user name: 29/29 on all six
2026-10-06 23:08:14 +08:00
parent 9862c29f58
commit 2a43ae6622
84 changed files with 9784 additions and 1083 deletions
@@ -12,12 +12,7 @@
 #include <string.h>
 #include <sys/stat.h>

-#ifndef _WIN32
-#include <dirent.h>
-#include <fnmatch.h>
 #include <regex.h>
-#include <unistd.h>
-#endif

 #include "cst_tools.h"
 #include "cst_util.h"
@@ -154,6 +149,9 @@ int cst_tool_read_file2(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
  FILE *fp = fopen(full, "rb");
  if (!fp)
    return tool_set_resultf(r, 1, "error: cannot open file: %.200s", path);
+  char bom[3];
+  if (fread(bom, 1, 3, fp) != 3 || !cst_bom_len(bom, 3))
+    rewind(fp);

  long offset = tool_arg_long(c, "offset", 0);
  long limit = tool_arg_long(c, "limit", 0);
@@ -178,6 +176,8 @@ int cst_tool_read_file2(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
      return tool_set_resultf(r, 1, "error: %.200s is a binary file (%zu bytes)",
                              path, n);
    }
+    cst_strip_bom(buf, &n);
+    n = cst_crlf_to_lf(buf, n); /* edits match LF text and keep CRLF */
    cst_tool_result_free(r);
    r->out = buf;
    r->out_len = n;
@@ -217,6 +217,9 @@ int cst_tool_read_file2(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
    }
    if (numbers)
      cst_buf_printf(&b, "%6ld|", lineno);
+    if (line.len >= 2 && line.buf[line.len - 2] == '\r' &&
+        line.buf[line.len - 1] == '\n')
+      line.buf[--line.len - 1] = '\n';
    cst_buf_append(&b, line.buf, line.len);
    shown++;
  }
@@ -442,6 +445,14 @@ static int plan_edit(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
        nwn = n2n;
      }
    }
+  } else if (cnt && memchr(nw, '\n', nwn) &&
+             cst_text_is_crlf(pl->oldc, pl->olen)) {
+    size_t n2n;
+    n2 = to_crlf(nw, nwn, &n2n);
+    if (n2) {
+      nw = n2;
+      nwn = n2n;
+    }
  }
  int rc = 0;
  if (cnt == 0) {
@@ -539,7 +550,6 @@ int cst_preview_edit_file(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
 /* --- delete_file / move_file --------------------------------------------- */

 static int reject_symlink(const cst_tool_ctx_t *ctx, const char *path) {
-#ifndef _WIN32
  char root[OS_PATH_MAX], joined[OS_PATH_MAX];
  if (os_realpath(ctx->workspace_root, root, sizeof(root)) != 0)
    return 1;
@@ -547,13 +557,9 @@ static int reject_symlink(const cst_tool_ctx_t *ctx, const char *path) {
    snprintf(joined, sizeof(joined), "%s", path);
  else if (os_path_join(joined, sizeof(joined), root, path) != 0)
    return 1;
-  struct stat st;
-  if (lstat(joined, &st) == 0 && S_ISLNK(st.st_mode))
+  os_stat_t st;
+  if (os_lstat(joined, &st) == 0 && st.is_link)
    return 1;
-#else
-  (void)ctx;
-  (void)path;
-#endif
  return 0;
 }

@@ -672,8 +678,6 @@ int cst_tool_move_file(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,

 /* --- directory walking (list_dir / glob / grep) -------------------------- */

-#ifndef _WIN32
-
 static const char *const SKIP_DIRS[] = {
    ".git", ".hg", ".svn", "node_modules", ".agent", "__pycache__",
    ".venv", ".cache", ".idea", ".vscode", ".mypy_cache", ".pytest_cache"};
@@ -703,7 +707,7 @@ static void ignore_load(const char *root, ignore_t *ig) {
  char p[OS_PATH_MAX];
  if (os_path_join(p, sizeof(p), root, ".gitignore") != 0)
    return;
-  FILE *fp = fopen(p, "r");
+  FILE *fp = fopen(p, "rb");
  if (!fp)
    return;
  char line[512];
@@ -749,7 +753,7 @@ static int ignored(const ignore_t *ig, const char *rel, const char *name,
    const ign_rule_t *r = &ig->v[i];
    if (r->dir_only && !isdir)
      continue;
-    if (fnmatch(r->pat, r->anchored ? rel : name, 0) == 0)
+    if (cst_fnmatch(r->pat, r->anchored ? rel : name, 0) == 0)
      return 1;
  }
  return 0;
@@ -758,7 +762,7 @@ static int ignored(const ignore_t *ig, const char *rel, const char *name,
 typedef struct walker walker_t;
 /* Return non-zero to stop the walk. */
 typedef int (*walk_fn)(walker_t *w, const char *rel, const char *full,
-                       const struct stat *st, int depth);
+                       const os_stat_t *st, int depth);

 struct walker {
  const char *root; /* workspace realpath */
@@ -781,15 +785,13 @@ static void walk_dir(walker_t *w, const char *rel, int depth) {
    snprintf(dirfull, sizeof(dirfull), "%s/%s", w->root, rel);
  else
    snprintf(dirfull, sizeof(dirfull), "%s", w->root);
-  DIR *d = opendir(dirfull);
+  os_dir_t *d = os_dir_open(dirfull);
  if (!d)
    return;
  char **names = NULL;
  int n = 0, cap = 0;
-  struct dirent *de;
-  while ((de = readdir(d)) != NULL) {
-    if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
-      continue;
+  const os_dirent_t *de;
+  while ((de = os_dir_next(d)) != NULL) {
    if (n == cap) {
      cap = cap ? cap * 2 : 32;
      char **nv = realloc(names, (size_t)cap * sizeof(char *));
@@ -797,11 +799,11 @@ static void walk_dir(walker_t *w, const char *rel, int depth) {
        break;
      names = nv;
    }
-    names[n] = strdup(de->d_name);
+    names[n] = strdup(de->name);
    if (names[n])
      n++;
  }
-  closedir(d);
+  os_dir_close(d);
  qsort(names, (size_t)n, sizeof(char *), name_cmp);
  for (int i = 0; i < n && !w->stop; i++) {
    char crel[OS_PATH_MAX], cfull[OS_PATH_MAX];
@@ -810,10 +812,10 @@ static void walk_dir(walker_t *w, const char *rel, int depth) {
    int k2 = snprintf(cfull, sizeof(cfull), "%s/%s", w->root, crel);
    if (k1 >= (int)sizeof(crel) || k2 >= (int)sizeof(cfull))
      continue; /* path too long */
-    struct stat st;
-    if (lstat(cfull, &st) != 0)
+    os_stat_t st;
+    if (os_lstat(cfull, &st) != 0)
      continue;
-    int isdir = S_ISDIR(st.st_mode);
+    int isdir = st.is_dir;
    if (ignored(&w->ig, crel, names[i], isdir))
      continue;
    if (w->fn(w, crel, cfull, &st, depth))
@@ -861,21 +863,18 @@ typedef struct {

 /* Entries directly inside a directory (for the depth boundary). */
 static int dir_entries(const char *full) {
-  DIR *d = opendir(full);
+  os_dir_t *d = os_dir_open(full);
  if (!d)
    return -1;
  int n = 0;
-  struct dirent *de;
-  while ((de = readdir(d)) != NULL) {
-    if (strcmp(de->d_name, ".") != 0 && strcmp(de->d_name, "..") != 0)
-      n++;
-  }
-  closedir(d);
+  while (os_dir_next(d))
+    n++;
+  os_dir_close(d);
  return n;
 }

 static int list_cb(walker_t *w, const char *rel, const char *full,
-                   const struct stat *st, int depth) {
+                   const os_stat_t *st, int depth) {
  list_ud_t *u = w->ud;
  if (u->count >= LIST_MAX) {
    if (++u->more >= LIST_COUNT_MAX) {
@@ -888,7 +887,7 @@ static int list_cb(walker_t *w, const char *rel, const char *full,
  name = name ? name + 1 : rel;
  for (int i = 0; i < depth; i++)
    cst_buf_puts(&u->b, "  ");
-  if (S_ISDIR(st->st_mode)) {
+  if (st->is_dir) {
    int k = -1;
    if (skip_dir_name(name))
      cst_buf_printf(&u->b, "%s/  (skipped)\n", name);
@@ -897,10 +896,10 @@ static int list_cb(walker_t *w, const char *rel, const char *full,
                     k == 1 ? "y" : "ies");
    else
      cst_buf_printf(&u->b, "%s/%s\n", name, k == 0 ? "  (empty)" : "");
-  } else if (S_ISLNK(st->st_mode))
+  } else if (st->is_link)
    cst_buf_printf(&u->b, "%s@\n", name);
  else
-    cst_buf_printf(&u->b, "%s  (%lld B)\n", name, (long long)st->st_size);
+    cst_buf_printf(&u->b, "%s  (%lld B)\n", name, st->size);
  u->count++;
  return 0;
 }
@@ -974,7 +973,7 @@ static int seg_match(const char *pat, const char *path) {
  pseg[pl] = 0;
  memcpy(sseg, path, sl);
  sseg[sl] = 0;
-  if (fnmatch(pseg, sseg, FNM_PERIOD) != 0)
+  if (cst_fnmatch(pseg, sseg, CST_FNM_PERIOD) != 0)
    return 0;
  if (!ps)
    return !ss;
@@ -994,11 +993,11 @@ typedef struct {
 } glob_ud_t;

 static int glob_cb(walker_t *w, const char *rel, const char *full,
-                   const struct stat *st, int depth) {
+                   const os_stat_t *st, int depth) {
  (void)full;
  (void)depth;
  glob_ud_t *u = w->ud;
-  if (!S_ISREG(st->st_mode))
+  if (!st->is_reg)
    return 0;
  const char *sub = rel;
  size_t bl = strlen(u->base);
@@ -1009,7 +1008,7 @@ static int glob_cb(walker_t *w, const char *rel, const char *full,
  }
  const char *name = strrchr(sub, '/');
  name = name ? name + 1 : sub;
-  int ok = u->basename_only ? fnmatch(u->pattern, name, FNM_PERIOD) == 0
+  int ok = u->basename_only ? cst_fnmatch(u->pattern, name, CST_FNM_PERIOD) == 0
                            : seg_match(u->pattern, sub);
  if (!ok)
    return 0;
@@ -1025,7 +1024,7 @@ static int glob_cb(walker_t *w, const char *rel, const char *full,
      return 1;
  }
  u->hits[u->n] = strdup(rel);
-  u->mtimes[u->n] = (double)st->st_mtime;
+  u->mtimes[u->n] = st->mtime;
  if (u->hits[u->n])
    u->n++;
  if (u->n >= GLOB_MAX * 20) {
@@ -1109,11 +1108,11 @@ typedef struct {
 } files_ud_t;

 static int files_cb(walker_t *w, const char *rel, const char *full,
-                    const struct stat *st, int depth) {
+                    const os_stat_t *st, int depth) {
  (void)full;
  (void)depth;
  files_ud_t *u = w->ud;
-  if (!S_ISREG(st->st_mode))
+  if (!st->is_reg)
    return 0;
  if (u->n >= u->max)
    return 1;
@@ -1275,16 +1274,16 @@ static void grep_file(grep_ud_t *u, const char *rel, const char *full) {
 }

 static int grep_cb(walker_t *w, const char *rel, const char *full,
-                   const struct stat *st, int depth) {
+                   const os_stat_t *st, int depth) {
  (void)depth;
  grep_ud_t *u = w->ud;
-  if (!S_ISREG(st->st_mode) || st->st_size > READ_FILE_MAX)
+  if (!st->is_reg || st->size > READ_FILE_MAX)
    return 0;
  if (u->glob) {
    const char *name = strrchr(rel, '/');
    name = name ? name + 1 : rel;
-    int ok = strchr(u->glob, '/') ? fnmatch(u->glob, rel, 0) == 0
-                                  : fnmatch(u->glob, name, 0) == 0;
+    int ok = strchr(u->glob, '/') ? cst_fnmatch(u->glob, rel, 0) == 0
+                                  : cst_fnmatch(u->glob, name, 0) == 0;
    if (!ok)
      return 0;
  }
@@ -1391,32 +1390,4 @@ int cst_tool_grep(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
  return tool_take_buf(r, &out, 0);
 }

-#else /* _WIN32: directory tools are not implemented natively */

-int cst_tool_list_dir(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
-                      cst_tool_result_t *r) {
-  (void)ctx;
-  (void)c;
-  return tool_set_result(r, "error: list_dir is not available on Windows; "
-                            "use run_shell", 1);
-}
-int cst_workspace_files(const char *ws, char ***out, int max) {
-  (void)ws;
-  (void)max;
-  *out = NULL;
-  return 0;
-}
-int cst_tool_glob(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
-                  cst_tool_result_t *r) {
-  (void)ctx;
-  (void)c;
-  return tool_set_result(r, "error: glob is not available on Windows", 1);
-}
-int cst_tool_grep(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
-                  cst_tool_result_t *r) {
-  (void)ctx;
-  (void)c;
-  return tool_set_result(r, "error: grep is not available on Windows", 1);
-}
-
-#endif
@@ -1,4 +1,5 @@
 #include "cst_agent.h"
+#include "cst_util.h"

 #include <stdio.h>
 #include <stdlib.h>
@@ -129,9 +130,9 @@ static int parse_hunk_header(const char *p, size_t n, long *a, long *b,
  return -1;
 }

-int cst_patch_apply_buffer(const char *orig, size_t olen, const char *hunks,
-                           size_t hlen, char **out, size_t *out_len,
-                           int *nhunks, char *err, size_t err_sz) {
+static int patch_core(const char *orig, size_t olen, const char *hunks,
+                      size_t hlen, char **out, size_t *out_len, int *nhunks,
+                      char *err, size_t err_sz) {
  lines_t file = {0}, hl = {0};
  lines_t old_seq = {0}, new_seq = {0};
  int rc = -1;
@@ -315,3 +316,47 @@ done:
  free(new_seq.v);
  return rc;
 }
+
+/* A CRLF file is patched as LF (the model's hunks usually are) and written
+ * back as CRLF; a BOM is kept out of the first line's context. */
+int cst_patch_apply_buffer(const char *orig, size_t olen, const char *hunks,
+                           size_t hlen, char **out, size_t *out_len,
+                           int *nhunks, char *err, size_t err_sz) {
+  size_t bom = orig ? cst_bom_len(orig, olen) : 0;
+  int crlf = orig && cst_text_is_crlf(orig, olen);
+  if (!bom && !crlf)
+    return patch_core(orig, olen, hunks, hlen, out, out_len, nhunks, err,
+                      err_sz);
+  if (out)
+    *out = NULL;
+  char *o = malloc(olen - bom + 1), *h = malloc(hlen + 1), *res = NULL;
+  size_t rlen = 0;
+  int rc = -1;
+  if (!o || !h) {
+    snprintf(err, err_sz, "oom");
+    goto done;
+  }
+  memcpy(o, orig + bom, olen - bom);
+  o[olen - bom] = 0;
+  size_t on = olen - bom, hn = hlen;
+  if (hlen)
+    memcpy(h, hunks, hlen);
+  h[hlen] = 0;
+  if (crlf) {
+    on = cst_crlf_to_lf(o, on);
+    hn = cst_crlf_to_lf(h, hn);
+  }
+  rc = patch_core(o, on, h, hn, &res, &rlen, nhunks, err, err_sz);
+  if (rc == 0) {
+    *out = cst_text_restyle(res, rlen, crlf, bom != 0, out_len);
+    if (!*out) {
+      snprintf(err, err_sz, "oom");
+      rc = -1;
+    }
+  }
+done:
+  free(res);
+  free(o);
+  free(h);
+  return rc;
+}
@@ -6,9 +6,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-#ifndef _WIN32
 #include <pthread.h>
-#endif

 #include "cJSON.h"
 #include "cst_tools.h"
@@ -182,14 +180,9 @@ const char *cst_tool_profile_name(int p) {
 #define DYN_MAX 256
 static cst_tool_def_t *g_dyn[DYN_MAX];
 static atomic_int g_ndyn;
-#ifndef _WIN32
 static pthread_mutex_t g_dyn_mu = PTHREAD_MUTEX_INITIALIZER;
 #define DYN_LOCK() pthread_mutex_lock(&g_dyn_mu)
 #define DYN_UNLOCK() pthread_mutex_unlock(&g_dyn_mu)
-#else
-#define DYN_LOCK() ((void)0)
-#define DYN_UNLOCK() ((void)0)
-#endif

 int cst_tool_count(void) { return NTOOLS + atomic_load(&g_ndyn); }

@@ -238,6 +238,22 @@ static int write_content(const ctl_tool_call_t *call, const char **content,
  return *content ? 0 : -1;
 }

+/* Overwriting a CRLF and/or BOM file with plain LF text keeps the file's
+ * style: malloc'd restyled copy, or NULL to write content as given. */
+static char *keep_text_style(const char *oldc, size_t olen, const char *s,
+                             size_t n, size_t *out_len) {
+  if (!oldc)
+    return NULL;
+  int has_crlf = 0;
+  for (size_t i = 1; i < n && !has_crlf; i++)
+    has_crlf = s[i] == '\n' && s[i - 1] == '\r';
+  int crlf = !has_crlf && cst_text_is_crlf(oldc, olen);
+  int bom = cst_bom_len(oldc, olen) && !cst_bom_len(s, n);
+  if (!crlf && !bom)
+    return NULL;
+  return cst_text_restyle(s, n, crlf, 1, out_len);
+}
+
 int cst_preview_write_file(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
                           char **out) {
  const char *path = tool_arg(c, "path");
@@ -252,11 +268,18 @@ int cst_preview_write_file(const cst_tool_ctx_t *ctx, const ctl_tool_call_t *c,
  char *oldc = NULL;
  if (cst_sandbox_resolve(ctx->workspace_root, path, full, sizeof(full)) == 0)
    oldc = tool_slurp(full, &olen, 0);
+  size_t sn = 0;
+  char *styled = keep_text_style(oldc, olen, content, n, &sn);
+  if (styled) {
+    content = styled;
+    n = sn;
+  }
  cst_buf_t b = {0};
  cst_buf_printf(&b, "%s %s (%zu bytes)\n", oldc ? "overwrite" : "create",
                 path, n);
  tool_append_diff(&b, path, oldc, olen, content, n);
  free(oldc);
+  free(styled);
  *out = cst_buf_steal(&b);
  return 0;
 }
@@ -276,14 +299,22 @@ int cst_tool_write_file(const char *workspace_root, const ctl_tool_call_t *call,

  size_t olen = 0;
  char *oldc = tool_slurp(full, &olen, 0);
+  size_t sn = 0;
+  char *styled = keep_text_style(oldc, olen, content, n, &sn);
+  if (styled) {
+    content = styled;
+    n = sn;
+  }
  if (tool_write_whole(full, content, n) != 0) {
    free(oldc);
+    free(styled);
    return tool_set_resultf(r, 1, "error: cannot write file: %.200s", path);
  }
  cst_buf_t b = {0};
  cst_buf_printf(&b, "wrote %s (%zu bytes)\n", path, n);
  tool_append_diff(&b, path, oldc, olen, content, n);
  free(oldc);
+  free(styled);
  return tool_take_buf(r, &b, 0);
 }