Add GitHub action and Makefile target for Windows installer

.github/workflows/installer-for-windows.yml

@@ -0,0 +1,27 @@
+# GitHub actions - Create Tesseract installer for Windows
+
+name: Cross build for Windows
+
+on:
+  # Trigger workflow in GitHub web frontend or from API.
+  workflow_dispatch:
+    inputs:
+      targets:
+        description: 'Target operating system'
+        required: true
+        default: 'Windows (64 bit)'
+        type: choice
+        options:
+          - 'Windows (64 bit)'
+
+jobs:
+  build64:
+    runs-on: [ubuntu-24.04]
+    steps:
+    - uses: actions/checkout@v4
+    - name: Build Tesseract installer (64 bit)
+      run: nsis/build.sh x86_64
+    - uses: actions/upload-artifact@v4
+      with:
+        name: Tesseract Installer for Windows (64 bit)
+        path: dist

.github/workflows/pkg-config-crosswrapper

@@ -0,0 +1,54 @@
+#! /bin/sh
+# pkg-config wrapper for cross-building
+# Sets pkg-config search path to search multiarch and historical cross-compiling paths.
+
+# If the user has already set PKG_CONFIG_LIBDIR, believe it (even if empty):
+# it's documented to be an override
+if [ x"${PKG_CONFIG_LIBDIR+set}" = x ]; then
+  # GNU triplet for the compiler, e.g. i486-linux-gnu for Debian i386,
+  # i686-linux-gnu for Ubuntu i386
+  basename="$(basename "$0")"
+  triplet="${basename%-pkg-config}"
+  # Normalized multiarch path if any, e.g. i386-linux-gnu for i386
+  dpkg-architecture >/dev/null 2>&1
+  if [ "$?" != 0 ]; then
+      # dpkg-architecture is missing.
+      echo "Please install dpkg-dev to use pkg-config when cross-building" >&2
+      exit 1
+  fi
+  multiarch="$(dpkg-architecture -t"${triplet}" -qDEB_HOST_MULTIARCH 2>/dev/null)"
+  # Native multiarch path
+  native_multiarch="$(cat /usr/lib/pkg-config.multiarch)"
+
+  # This can be used for native builds as well, in that case, just exec pkg-config "$@" directly.
+  if [ "$native_multiarch" = "$multiarch" ]; then
+     exec pkg-config "$@"
+  fi
+
+  PKG_CONFIG_LIBDIR="/usr/local/${triplet}/lib/pkgconfig"
+  # For a native build we would also want to append /usr/local/lib/pkgconfig
+  # at this point; but this is a cross-building script, so don't
+  PKG_CONFIG_LIBDIR="$PKG_CONFIG_LIBDIR:/usr/local/share/pkgconfig"
+
+  if [ -n "$multiarch" ]; then
+    PKG_CONFIG_LIBDIR="/usr/local/lib/${multiarch}/pkgconfig:$PKG_CONFIG_LIBDIR"
+    PKG_CONFIG_LIBDIR="$PKG_CONFIG_LIBDIR:/usr/lib/${multiarch}/pkgconfig"
+  fi
+
+  PKG_CONFIG_LIBDIR="$PKG_CONFIG_LIBDIR:/usr/${triplet}/lib/pkgconfig"
+  # For a native build we would also want to append /usr/lib/pkgconfig
+  # at this point; but this is a cross-building script, so don't
+  # If you want to allow use of un-multiarched -dev packages for crossing
+  # (at the risk of finding build-arch stuff you didn't want, if not in a clean chroot)
+  # Uncomment the next line:
+  # PKG_CONFIG_LIBDIR="$PKG_CONFIG_LIBDIR:/usr/lib/pkgconfig"
+  # ... but on Ubuntu we rely cross-building with non-multiarch libraries:
+  if dpkg-vendor --derives-from Ubuntu; then
+      PKG_CONFIG_LIBDIR="$PKG_CONFIG_LIBDIR:/usr/lib/pkgconfig"
+  fi
+  PKG_CONFIG_LIBDIR="$PKG_CONFIG_LIBDIR:/usr/share/pkgconfig"
+
+  export PKG_CONFIG_LIBDIR
+fi
+
+exec pkg-config "$@"

Makefile.am

@@ -6,6 +6,9 @@ ACLOCAL_AMFLAGS = -I m4
 CLEANFILES =
 
 SUBDIRS = . tessdata
+if MINGW
+SUBDIRS += nsis
+endif
 
 EXTRA_DIST = README.md LICENSE
 EXTRA_DIST += aclocal.m4 config configure.ac autogen.sh
@@ -50,6 +53,11 @@ doc-pack: doc
 doc-clean:
 	rm -rf $(top_builddir)/doc/html/*
 
+if MINGW
+winsetup: training ScrollView.jar
+	@cd "$(top_builddir)/nsis" && $(MAKE) winsetup
+endif
+
 pkgconfigdir = $(libdir)/pkgconfig
 pkgconfig_DATA = tesseract.pc
 

configure.ac

@@ -72,6 +72,7 @@ AC_CONFIG_HEADERS([include/config_auto.h:config/config.h.in])
 
 # default conditional
 AM_CONDITIONAL([T_WIN], false)
+AM_CONDITIONAL([MINGW], false)
 AM_CONDITIONAL([GRAPHICS_DISABLED], false)
 AC_SUBST([AM_CPPFLAGS])
 
@@ -87,7 +88,9 @@ AM_SILENT_RULES([yes])
 AC_CANONICAL_HOST
 case "${host_os}" in
     mingw*)
+        AC_DEFINE_UNQUOTED([MINGW], 1, [This is a MinGW system])
         AM_CONDITIONAL([T_WIN], true)
+        AM_CONDITIONAL([MINGW], true)
         AM_CONDITIONAL([ADD_RT], false)
         AC_SUBST([AM_LDFLAGS], ['-no-undefined'])
         ;;
@@ -559,6 +562,7 @@ AC_CONFIG_FILES([java/com/google/Makefile])
 AC_CONFIG_FILES([java/com/google/scrollview/Makefile])
 AC_CONFIG_FILES([java/com/google/scrollview/events/Makefile])
 AC_CONFIG_FILES([java/com/google/scrollview/ui/Makefile])
+AC_CONFIG_FILES([nsis/Makefile])
 AC_OUTPUT
 
 # Final message

nsis/Makefile.am

@@ -0,0 +1,22 @@
+AUTOMAKE_OPTIONS = subdir-objects
+
+all:
+
+if MINGW
+
+gitrev="$(shell git --git-dir=${abs_top_srcdir}/.git --work-tree=${abs_top_srcdir} describe --always --tags | sed s/^v//)"
+
+.PHONY: winsetup
+
+Plugins/x86-unicode/INetC.dll:
+	curl -O https://nsis.sourceforge.io/mediawiki/images/c/c9/Inetc.zip
+	unzip Inetc.zip $@
+
+winpath.exe: winpath.cpp
+	x86_64-w64-mingw32-g++ -Os -o $@ $<
+	x86_64-w64-mingw32-strip --strip-unneeded $@
+
+winsetup: Plugins/x86-unicode/INetC.dll winpath.exe
+	makensis -DCROSSBUILD -DSHARED -DSIGNCODE=$(SIGNCODE) -DSRCDIR=$(top_srcdir) -DVERSION=${gitrev} $(shell test "$(host_cpu)" = x86_64 && echo "-DW64") -NOCD $(top_srcdir)/nsis/tesseract.nsi
+
+endif

nsis/build.sh

@@ -0,0 +1,118 @@
+#!/bin/bash
+
+# GitHub actions - Create Tesseract installer for Windows
+
+# Author: Stefan Weil (2010-2024)
+
+set -e
+set -x
+
+LANG=C.UTF-8
+
+ARCH=$1
+
+if [ "$ARCH" = "i686" ]; then
+  MINGW=/mingw32
+else
+  ARCH=x86_64
+  MINGW=/mingw64
+fi
+
+ROOTDIR=$PWD
+DISTDIR=$ROOTDIR/dist
+HOST=$ARCH-w64-mingw32
+TAG=$(cat VERSION).$(date +%Y%m%d)
+BUILDDIR=bin/ndebug/$HOST-$TAG
+PKG_ARCH=mingw-w64-${ARCH/_/-}
+
+# Install packages.
+sudo apt-get update --quiet
+sudo apt-get install --assume-yes --no-install-recommends --quiet \
+  asciidoc xsltproc docbook-xml docbook-xsl \
+  automake dpkg-dev libtool pkg-config default-jdk-headless \
+  mingw-w64-tools nsis g++-"$PKG_ARCH" \
+  makepkg pacman-package-manager
+
+# Install pacman-package-manager and its dependencies (from Ubuntu 22.10).
+# sudo curl -Os http://de.archive.ubuntu.com/ubuntu/pool/universe/p/pacman-package-manager/pacman-package-manager_6.0.1-4_amd64.deb
+# sudo curl -Os http://de.archive.ubuntu.com/ubuntu/pool/universe/p/pacman-package-manager/libalpm13_6.0.1-4_amd64.deb
+# sudo curl -Os http://de.archive.ubuntu.com/ubuntu/pool/universe/p/pacman-package-manager/makepkg_6.0.1-4_amd64.deb
+# sudo dpkg -i *.deb || true
+# sudo apt-get install --fix-broken --assume-yes --no-install-recommends --quiet
+
+# Configure pacman.
+
+# Enable mirrorlist.
+sudo sed -Ei 's/^#.*(Include.*mirrorlist)/\1/' /etc/pacman.conf
+(
+# Add msys key for pacman.
+cd /usr/share/keyrings
+sudo curl -Os https://raw.githubusercontent.com/msys2/MSYS2-keyring/master/msys2.gpg
+sudo curl -Os https://raw.githubusercontent.com/msys2/MSYS2-keyring/master/msys2-revoked
+sudo curl -Os https://raw.githubusercontent.com/msys2/MSYS2-keyring/master/msys2-trusted
+)
+(
+# Add active environments for pacman.
+# See https://www.msys2.org/docs/repos-mirrors/.
+sudo mkdir -p /etc/pacman.d
+cd /etc/pacman.d
+cat <<eod | sudo tee mirrorlist >/dev/null
+[mingw64]
+Include = /etc/pacman.d/mirrorlist.mingw
+eod
+sudo curl -O https://raw.githubusercontent.com/msys2/MSYS2-packages/master/pacman-mirrors/mirrorlist.mingw
+# sudo curl -O https://raw.githubusercontent.com/msys2/MSYS2-packages/master/pacman-mirrors/mirrorlist.msys
+)
+
+sudo pacman-key --init
+sudo pacman-key --populate msys2
+sudo pacman -Syu --noconfirm
+
+# Install required pacman packages.
+sudo pacman -S --noconfirm \
+ mingw-w64-x86_64-curl-winssl \
+ mingw-w64-x86_64-giflib \
+ mingw-w64-x86_64-icu \
+ mingw-w64-x86_64-leptonica \
+ mingw-w64-x86_64-libarchive \
+ mingw-w64-x86_64-libidn2 \
+ mingw-w64-x86_64-openjpeg2 \
+ mingw-w64-x86_64-openssl \
+ mingw-w64-x86_64-pango \
+ mingw-w64-x86_64-libpng \
+ mingw-w64-x86_64-libtiff \
+ mingw-w64-x86_64-libwebp
+
+sudo ln -sf "$PWD/.github/workflows/pkg-config-crosswrapper" "/usr/bin/$HOST-pkg-config"
+
+git config --global user.email "[email protected]"
+git config --global user.name "Stefan Weil"
+git tag -a "v$TAG" -m "Tesseract $TAG"
+
+# Run autogen.
+./autogen.sh
+
+# Build Tesseract installer.
+mkdir -p "$BUILDDIR" && cd "$BUILDDIR"
+
+# Run configure.
+PKG_CONFIG_PATH=$MINGW/lib/pkgconfig
+export PKG_CONFIG_PATH
+# Disable OpenMP (see https://github.com/tesseract-ocr/tesseract/issues/1662).
+../../../configure --disable-openmp --host="$HOST" --prefix="/usr/$HOST" \
+  CXX="$HOST-g++-posix" \
+  CXXFLAGS="-fno-math-errno -Wall -Wextra -Wpedantic -g -O2 -isystem $MINGW/include" \
+  LDFLAGS="-L$MINGW/lib"
+
+make all training
+MINGW_INSTALL=${PWD}${MINGW}
+make install-jars install training-install html prefix="$MINGW_INSTALL"
+test -d venv || python3 -m venv venv
+source venv/bin/activate
+pip install pefile
+mkdir -p dll
+ln -sv $("$ROOTDIR/nsis/find_deps.py" "$MINGW_INSTALL"/bin/*.exe "$MINGW_INSTALL"/bin/*.dll) dll/
+make winsetup prefix="$MINGW_INSTALL"
+
+# Copy result for upload.
+mkdir -p "$DISTDIR" && cp nsis/tesseract-ocr-w*-setup-*.exe "$DISTDIR"

nsis/find_deps.py

@@ -0,0 +1,64 @@
+#!/usr/bin/env python3
+#
+# Copyright (C) 2024 Stefan Weil
+#
+# SPDX-License-Identifier: MIT
+#
+# Find the DLL files which are required for a given set of
+# Windows executables and libraries.
+
+import argparse
+import os
+import pefile
+
+VERBOSE = False
+
+def find_dependencies(binary, search_path, analyzed_deps):
+    pe = pefile.PE(binary)
+    pe.parse_data_directories()
+    if VERBOSE:
+        print(f'{binary}:')
+    # print(pe.dump_info())
+
+    for entry in pe.DIRECTORY_ENTRY_IMPORT:
+        name = entry.dll.decode('utf-8')
+        if name in analyzed_deps:
+            if VERBOSE:
+                print(f'skip {name} (already analyzed)')
+            continue
+        analyzed_deps.add(name)
+        fullpath = os.path.join(search_path, name)
+        if not os.path.exists(fullpath):
+            # Not found, maybe system DLL. Skip it.
+            if VERBOSE:
+                print(f'skip {name} (not found, maybe system DLL)')
+            continue
+        print(fullpath)
+        analyzed_deps = find_dependencies(fullpath, search_path, analyzed_deps)
+
+    return analyzed_deps
+
+def main():
+    """
+    Command-line interface for universal dependency scanner.
+    """
+
+    parser = argparse.ArgumentParser(description='Find and copy DLL dependencies')
+    parser.add_argument('files', nargs='+', help='Paths to executable or library files')
+    parser.add_argument('--dlldir', dest='dlldir', default='/mingw64/bin/',
+                        help='path to dll files')
+
+    args = parser.parse_args()
+
+    # try:
+    # Find dependencies
+    analyzed_deps = set()
+    for binary in args.files:
+        if True:
+            analyzed_deps = find_dependencies(binary, args.dlldir, analyzed_deps)
+        # except:
+        #    print(f'error: failed to find dependencies for {binary}')
+
+
+if __name__ == '__main__':
+    main()