mtkports/www-client/netsurf/compat/iconv.c

/*
 * iconv() for the MontaukOS NetSurf port, implemented over libparserutils'
 * charset codecs. See iconv.h for why.
 *
 * Structure: one codec decodes the source charset to UCS-4, a second encodes
 * UCS-4 to the destination charset. iconv() pumps a small stack buffer of
 * UCS-4 between them so neither side needs an unbounded intermediate.
 */

#include <errno.h>
#include <stdlib.h>
#include <string.h>

#include <parserutils/charset/codec.h>
#include <parserutils/errors.h>

#include "iconv.h"

/* UCS-4 code points staged between decode and encode, per inner iteration. */
#define STAGE_POINTS 64

struct montauk_iconv {
	parserutils_charset_codec *from;	/* source charset -> UCS-4 */
	parserutils_charset_codec *to;		/* UCS-4 -> dest charset  */
};

/*
 * Strip glibc's "//TRANSLIT" / "//IGNORE" suffixes, which libparserutils does
 * not understand. Returns a malloc'd bare charset name.
 */
static char *strip_suffix(const char *name)
{
	const char *sep;
	size_t len;
	char *out;

	if (name == NULL)
		return NULL;

	sep = strstr(name, "//");
	len = (sep != NULL) ? (size_t) (sep - name) : strlen(name);

	out = malloc(len + 1);
	if (out == NULL)
		return NULL;

	memcpy(out, name, len);
	out[len] = '\0';
	return out;
}

static parserutils_charset_codec *make_codec(const char *name)
{
	parserutils_charset_codec *codec = NULL;
	parserutils_charset_codec_optparams params;
	char *bare;

	bare = strip_suffix(name);
	if (bare == NULL)
		return NULL;

	if (parserutils_charset_codec_create(bare, &codec) != PARSERUTILS_OK) {
		free(bare);
		return NULL;
	}
	free(bare);

	/*
	 * LOOSE: substitute unrepresentable characters instead of aborting.
	 * NetSurf renders best-effort; a hard failure mid-document is worse
	 * than a replacement character.
	 */
	params.error_mode.mode = PARSERUTILS_CHARSET_CODEC_ERROR_LOOSE;
	(void) parserutils_charset_codec_setopt(codec,
			PARSERUTILS_CHARSET_CODEC_ERROR_MODE, &params);

	return codec;
}

iconv_t iconv_open(const char *tocode, const char *fromcode)
{
	struct montauk_iconv *cd;

	cd = calloc(1, sizeof(*cd));
	if (cd == NULL) {
		errno = ENOMEM;
		return (iconv_t) -1;
	}

	cd->from = make_codec(fromcode);
	cd->to = make_codec(tocode);

	if (cd->from == NULL || cd->to == NULL) {
		if (cd->from != NULL)
			parserutils_charset_codec_destroy(cd->from);
		if (cd->to != NULL)
			parserutils_charset_codec_destroy(cd->to);
		free(cd);
		errno = EINVAL;		/* unsupported conversion */
		return (iconv_t) -1;
	}

	return (iconv_t) cd;
}

int iconv_close(iconv_t handle)
{
	struct montauk_iconv *cd = (struct montauk_iconv *) handle;

	if (cd == NULL || handle == (iconv_t) -1) {
		errno = EBADF;
		return -1;
	}

	parserutils_charset_codec_destroy(cd->from);
	parserutils_charset_codec_destroy(cd->to);
	free(cd);
	return 0;
}

size_t iconv(iconv_t handle, char **inbuf, size_t *inbytesleft,
	     char **outbuf, size_t *outbytesleft)
{
	struct montauk_iconv *cd = (struct montauk_iconv *) handle;
	size_t converted = 0;

	if (cd == NULL || handle == (iconv_t) -1) {
		errno = EBADF;
		return (size_t) -1;
	}

	/* Reset request: iconv(cd, NULL, NULL, ...) returns to initial state. */
	if (inbuf == NULL || *inbuf == NULL) {
		parserutils_charset_codec_reset(cd->from);
		parserutils_charset_codec_reset(cd->to);
		return 0;
	}

	while (*inbytesleft > 0) {
		uint8_t stage[STAGE_POINTS * 4];
		const uint8_t *src = (const uint8_t *) *inbuf;
		size_t srclen = *inbytesleft;
		uint8_t *mid = stage;
		size_t midlen = sizeof(stage);
		size_t staged;
		const uint8_t *midsrc;
		size_t midsrclen;
		uint8_t *dst;
		size_t dstlen;
		parserutils_error perr;

		/* ---- source charset -> UCS-4 ---- */
		perr = parserutils_charset_codec_decode(cd->from,
				&src, &srclen, &mid, &midlen);

		staged = sizeof(stage) - midlen;

		if (staged == 0) {
			/* nothing decoded: classify why and stop */
			if (perr == PARSERUTILS_NEEDDATA)
				errno = EINVAL;	/* incomplete tail sequence */
			else if (perr == PARSERUTILS_INVALID)
				errno = EILSEQ;	/* malformed input */
			else if (perr == PARSERUTILS_NOMEM)
				errno = E2BIG;
			else
				errno = EILSEQ;
			return (size_t) -1;
		}

		/* ---- UCS-4 -> destination charset ---- */
		midsrc = stage;
		midsrclen = staged;
		dst = (uint8_t *) *outbuf;
		dstlen = *outbytesleft;

		perr = parserutils_charset_codec_encode(cd->to,
				&midsrc, &midsrclen, &dst, &dstlen);

		/*
		 * Advance the caller's input only by what was both decoded AND
		 * encoded. If the encoder ran out of output room mid-stage,
		 * the undrained UCS-4 must not be counted as consumed -- the
		 * caller will grow its buffer and call again.
		 */
		{
			size_t encoded_points = (staged - midsrclen) / 4;
			size_t consumed_in;

			*outbytesleft = dstlen;
			*outbuf = (char *) dst;

			if (midsrclen == 0) {
				/* whole stage drained: input consumed as decoded */
				consumed_in = (size_t) ((const char *) src - *inbuf);
			} else {
				/*
				 * Partial drain. Re-decode from the original
				 * position next time; consume nothing this
				 * round beyond what actually made it out.
				 */
				if (encoded_points == 0) {
					errno = E2BIG;
					return (size_t) -1;
				}
				consumed_in = 0;
			}

			if (consumed_in > 0) {
				*inbuf += consumed_in;
				*inbytesleft -= consumed_in;
			}

			converted += encoded_points;

			if (midsrclen != 0) {
				/* output was too small to take the rest */
				errno = E2BIG;
				return (size_t) -1;
			}
		}

		if (perr == PARSERUTILS_NOMEM) {
			errno = E2BIG;
			return (size_t) -1;
		}

		if (*outbytesleft == 0 && *inbytesleft > 0) {
			errno = E2BIG;
			return (size_t) -1;
		}
	}

	return converted;
}