637 строки
15 KiB
C
637 строки
15 KiB
C
/*****************************************************************************
|
|
* _ _ ____ _
|
|
* Project ___| | | | _ \| |
|
|
* / __| | | | |_) | |
|
|
* | (__| |_| | _ <| |___
|
|
* \___|\___/|_| \_\_____|
|
|
*
|
|
* The contents of this file are subject to the Mozilla Public License
|
|
* Version 1.0 (the "License"); you may not use this file except in
|
|
* compliance with the License. You may obtain a copy of the License at
|
|
* http://www.mozilla.org/MPL/
|
|
*
|
|
* Software distributed under the License is distributed on an "AS IS"
|
|
* basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See the
|
|
* License for the specific language governing rights and limitations
|
|
* under the License.
|
|
*
|
|
* The Original Code is Curl.
|
|
*
|
|
* The Initial Developer of the Original Code is Daniel Stenberg.
|
|
*
|
|
* Portions created by the Initial Developer are Copyright (C) 1998.
|
|
* All Rights Reserved.
|
|
*
|
|
* ------------------------------------------------------------
|
|
* Main author:
|
|
* - Daniel Stenberg <daniel@haxx.se>
|
|
*
|
|
* http://curl.haxx.se
|
|
*
|
|
* $Source$
|
|
* $Revision$
|
|
* $Date$
|
|
* $Author$
|
|
* $State$
|
|
* $Locker$
|
|
*
|
|
* ------------------------------------------------------------
|
|
****************************************************************************/
|
|
|
|
/*
|
|
Debug the form generator stand-alone by compiling this source file with:
|
|
|
|
gcc -DHAVE_CONFIG_H -I../ -g -D_FORM_DEBUG -o formdata -I../include formdata.c
|
|
|
|
run the 'formdata' executable and make sure the output is ok!
|
|
|
|
try './formdata "name=Daniel" "poo=noo" "foo=bar"' and similarly
|
|
|
|
*/
|
|
|
|
#include "setup.h"
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <stdarg.h>
|
|
|
|
#include <time.h>
|
|
|
|
#include <curl/curl.h>
|
|
#include "formdata.h"
|
|
|
|
#include "strequal.h"
|
|
|
|
/* Length of the random boundary string. The risk of this being used
|
|
in binary data is very close to zero, 64^32 makes
|
|
6277101735386680763835789423207666416102355444464034512896
|
|
combinations... */
|
|
#define BOUNDARY_LENGTH 32
|
|
|
|
/* What kind of Content-Type to use on un-specified files with unrecognized
|
|
extensions. */
|
|
#define HTTPPOST_CONTENTTYPE_DEFAULT "text/plain"
|
|
|
|
/* This is a silly duplicate of the function in main.c to enable this source
|
|
to compile stand-alone for better debugging */
|
|
static void GetStr(char **string,
|
|
char *value)
|
|
{
|
|
if(*string)
|
|
free(*string);
|
|
*string = strdup(value);
|
|
}
|
|
|
|
/***************************************************************************
|
|
*
|
|
* FormParse()
|
|
*
|
|
* Reads a 'name=value' paramter and builds the appropriate linked list.
|
|
*
|
|
* Specify files to upload with 'name=@filename'. Supports specified
|
|
* given Content-Type of the files. Such as ';type=<content-type>'.
|
|
*
|
|
* You may specify more than one file for a single name (field). Specify
|
|
* multiple files by writing it like:
|
|
*
|
|
* 'name=@filename,filename2,filename3'
|
|
*
|
|
* If you want content-types specified for each too, write them like:
|
|
*
|
|
* 'name=@filename;type=image/gif,filename2,filename3'
|
|
*
|
|
***************************************************************************/
|
|
|
|
int curl_formparse(char *input,
|
|
struct HttpPost **httppost,
|
|
struct HttpPost **last_post)
|
|
{
|
|
return FormParse(input, httppost, last_post);
|
|
}
|
|
|
|
#define FORM_FILE_SEPARATOR ','
|
|
#define FORM_TYPE_SEPARATOR ';'
|
|
|
|
int FormParse(char *input,
|
|
struct HttpPost **httppost,
|
|
struct HttpPost **last_post)
|
|
{
|
|
/* nextarg MUST be a string in the format 'name=contents' and we'll
|
|
build a linked list with the info */
|
|
char name[256];
|
|
char contents[4096]="";
|
|
char major[128];
|
|
char minor[128];
|
|
long flags = 0;
|
|
char *contp;
|
|
char *type = NULL;
|
|
char *prevtype = NULL;
|
|
char *sep;
|
|
char *sep2;
|
|
struct HttpPost *post;
|
|
struct HttpPost *subpost; /* a sub-node */
|
|
unsigned int i;
|
|
|
|
if(1 <= sscanf(input, "%255[^ =] = %4095[^\n]", name, contents)) {
|
|
/* the input was using the correct format */
|
|
contp = contents;
|
|
|
|
if('@' == contp[0]) {
|
|
/* we use the @-letter to indicate file name(s) */
|
|
|
|
flags = HTTPPOST_FILENAME;
|
|
contp++;
|
|
|
|
post=NULL;
|
|
|
|
do {
|
|
/* since this was a file, it may have a content-type specifier
|
|
at the end too */
|
|
|
|
sep=strchr(contp, FORM_TYPE_SEPARATOR);
|
|
sep2=strchr(contp, FORM_FILE_SEPARATOR);
|
|
|
|
/* pick the closest */
|
|
if(sep2 && (sep2 < sep)) {
|
|
sep = sep2;
|
|
|
|
/* no type was specified! */
|
|
}
|
|
if(sep) {
|
|
|
|
/* if we got here on a comma, don't do much */
|
|
if(FORM_FILE_SEPARATOR != *sep)
|
|
type = strstr(sep+1, "type=");
|
|
else
|
|
type=NULL;
|
|
|
|
*sep=0; /* terminate file name at separator */
|
|
|
|
if(type) {
|
|
type += strlen("type=");
|
|
|
|
if(2 != sscanf(type, "%127[^/]/%127[^,\n]",
|
|
major, minor)) {
|
|
fprintf(stderr, "Illegally formatted content-type field!\n");
|
|
return 2; /* illegal content-type syntax! */
|
|
}
|
|
/* now point beyond the content-type specifier */
|
|
sep = type + strlen(major)+strlen(minor)+1;
|
|
|
|
/* find the following comma */
|
|
sep=strchr(sep, FORM_FILE_SEPARATOR);
|
|
}
|
|
}
|
|
else {
|
|
type=NULL;
|
|
sep=strchr(contp, FORM_FILE_SEPARATOR);
|
|
}
|
|
if(sep) {
|
|
/* the next file name starts here */
|
|
*sep =0;
|
|
sep++;
|
|
}
|
|
if(!type) {
|
|
/*
|
|
* No type was specified, we scan through a few well-known
|
|
* extensions and pick the first we match!
|
|
*/
|
|
struct ContentType {
|
|
char *extension;
|
|
char *type;
|
|
};
|
|
static struct ContentType ctts[]={
|
|
{".gif", "image/gif"},
|
|
{".jpg", "image/jpeg"},
|
|
{".jpeg", "image/jpeg"},
|
|
{".txt", "text/plain"},
|
|
{".html", "text/plain"}
|
|
};
|
|
|
|
if(prevtype)
|
|
/* default to the previously set/used! */
|
|
type = prevtype;
|
|
else
|
|
/* It seems RFC1867 defines no Content-Type to default to
|
|
text/plain so we don't actually need to set this: */
|
|
type = HTTPPOST_CONTENTTYPE_DEFAULT;
|
|
|
|
for(i=0; i<sizeof(ctts)/sizeof(ctts[0]); i++) {
|
|
if(strlen(contp) >= strlen(ctts[i].extension)) {
|
|
if(strequal(contp +
|
|
strlen(contp) - strlen(ctts[i].extension),
|
|
ctts[i].extension)) {
|
|
type = ctts[i].type;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
/* we have a type by now */
|
|
}
|
|
|
|
if(NULL == post) {
|
|
/* For the first file name, we allocate and initiate the main list
|
|
node */
|
|
|
|
post = (struct HttpPost *)malloc(sizeof(struct HttpPost));
|
|
if(post) {
|
|
memset(post, 0, sizeof(struct HttpPost));
|
|
GetStr(&post->name, name); /* get the name */
|
|
GetStr(&post->contents, contp); /* get the contents */
|
|
post->flags = flags;
|
|
if(type) {
|
|
GetStr(&post->contenttype, type); /* get type */
|
|
prevtype=post->contenttype; /* point to the allocated string! */
|
|
}
|
|
/* make the previous point to this */
|
|
if(*last_post)
|
|
(*last_post)->next = post;
|
|
else
|
|
(*httppost) = post;
|
|
|
|
(*last_post) = post;
|
|
}
|
|
|
|
}
|
|
else {
|
|
/* we add a file name to the previously allocated node, known as
|
|
'post' now */
|
|
subpost =(struct HttpPost *)malloc(sizeof(struct HttpPost));
|
|
if(subpost) {
|
|
memset(subpost, 0, sizeof(struct HttpPost));
|
|
GetStr(&subpost->name, name); /* get the name */
|
|
GetStr(&subpost->contents, contp); /* get the contents */
|
|
subpost->flags = flags;
|
|
if(type) {
|
|
GetStr(&subpost->contenttype, type); /* get type */
|
|
prevtype=subpost->contenttype; /* point to the allocated string! */
|
|
}
|
|
/* now, point our 'more' to the original 'more' */
|
|
subpost->more = post->more;
|
|
|
|
/* then move the original 'more' to point to ourselves */
|
|
post->more = subpost;
|
|
}
|
|
}
|
|
contp = sep; /* move the contents pointer to after the separator */
|
|
} while(sep && *sep); /* loop if there's another file name */
|
|
}
|
|
else {
|
|
post = (struct HttpPost *)malloc(sizeof(struct HttpPost));
|
|
if(post) {
|
|
memset(post, 0, sizeof(struct HttpPost));
|
|
GetStr(&post->name, name); /* get the name */
|
|
if( contp[0]=='<' ) {
|
|
GetStr(&post->contents, contp+1); /* get the contents */
|
|
post->flags = HTTPPOST_READFILE;
|
|
}
|
|
else {
|
|
GetStr(&post->contents, contp); /* get the contents */
|
|
post->flags = 0;
|
|
}
|
|
|
|
/* make the previous point to this */
|
|
if(*last_post)
|
|
(*last_post)->next = post;
|
|
else
|
|
(*httppost) = post;
|
|
|
|
(*last_post) = post;
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
else {
|
|
fprintf(stderr, "Illegally formatted input field!\n");
|
|
return 1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
static int AddFormData(struct FormData **formp,
|
|
void *line,
|
|
long length)
|
|
{
|
|
struct FormData *newform = (struct FormData *)
|
|
malloc(sizeof(struct FormData));
|
|
newform->next = NULL;
|
|
|
|
/* we make it easier for plain strings: */
|
|
if(!length)
|
|
length = strlen((char *)line);
|
|
|
|
newform->line = (char *)malloc(length+1);
|
|
memcpy(newform->line, line, length+1);
|
|
newform->length = length;
|
|
newform->line[length]=0; /* zero terminate for easier debugging */
|
|
|
|
if(*formp) {
|
|
(*formp)->next = newform;
|
|
*formp = newform;
|
|
}
|
|
else
|
|
*formp = newform;
|
|
|
|
return length;
|
|
}
|
|
|
|
|
|
static int AddFormDataf(struct FormData **formp,
|
|
char *fmt, ...)
|
|
{
|
|
char s[4096];
|
|
va_list ap;
|
|
va_start(ap, fmt);
|
|
vsprintf(s, fmt, ap);
|
|
va_end(ap);
|
|
|
|
return AddFormData(formp, s, 0);
|
|
}
|
|
|
|
|
|
char *MakeFormBoundary(void)
|
|
{
|
|
char *retstring;
|
|
static int randomizer=0; /* this is just so that two boundaries within
|
|
the same form won't be identical */
|
|
int i;
|
|
|
|
static char table64[]=
|
|
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
|
|
|
|
retstring = (char *)malloc(BOUNDARY_LENGTH);
|
|
|
|
if(!retstring)
|
|
return NULL; /* failed */
|
|
|
|
srand(time(NULL)+randomizer++); /* seed */
|
|
|
|
strcpy(retstring, "curl"); /* bonus commercials 8*) */
|
|
|
|
for(i=4; i<(BOUNDARY_LENGTH-1); i++) {
|
|
retstring[i] = table64[rand()%64];
|
|
}
|
|
retstring[BOUNDARY_LENGTH-1]=0; /* zero terminate */
|
|
|
|
return retstring;
|
|
}
|
|
|
|
|
|
void FormFree(struct FormData *form)
|
|
{
|
|
struct FormData *next;
|
|
do {
|
|
next=form->next; /* the following form line */
|
|
free(form->line); /* free the line */
|
|
free(form); /* free the struct */
|
|
|
|
} while((form=next)); /* continue */
|
|
}
|
|
|
|
struct FormData *getFormData(struct HttpPost *post,
|
|
int *sizep)
|
|
{
|
|
struct FormData *form = NULL;
|
|
struct FormData *firstform;
|
|
|
|
struct HttpPost *file;
|
|
|
|
int size =0;
|
|
char *boundary;
|
|
char *fileboundary=NULL;
|
|
|
|
if(!post)
|
|
return NULL; /* no input => no output! */
|
|
|
|
boundary = MakeFormBoundary();
|
|
|
|
/* Make the first line of the output */
|
|
AddFormDataf(&form,
|
|
"Content-Type: multipart/form-data;"
|
|
" boundary=%s\r\n",
|
|
boundary);
|
|
/* we DO NOT count that line since that'll be part of the header! */
|
|
|
|
firstform = form;
|
|
|
|
do {
|
|
|
|
/* boundary */
|
|
size += AddFormDataf(&form, "\r\n--%s\r\n", boundary);
|
|
|
|
size += AddFormDataf(&form,
|
|
"Content-Disposition: form-data; name=\"%s\"",
|
|
post->name);
|
|
|
|
if(post->more) {
|
|
/* If used, this is a link to more file names, we must then do
|
|
the magic to include several files with the same field name */
|
|
|
|
fileboundary = MakeFormBoundary();
|
|
|
|
size += AddFormDataf(&form,
|
|
"\r\nContent-Type: multipart/mixed,"
|
|
" boundary=%s\r\n",
|
|
fileboundary);
|
|
}
|
|
|
|
file = post;
|
|
|
|
do {
|
|
if(post->more) {
|
|
/* if multiple-file */
|
|
size += AddFormDataf(&form,
|
|
"\r\n--%s\r\nContent-Disposition: attachment; filename=\"%s\"",
|
|
fileboundary, file->contents);
|
|
}
|
|
else if(post->flags & HTTPPOST_FILENAME) {
|
|
size += AddFormDataf(&form,
|
|
"; filename=\"%s\"",
|
|
post->contents);
|
|
}
|
|
|
|
if(file->contenttype) {
|
|
/* we have a specified type */
|
|
size += AddFormDataf(&form,
|
|
"\r\nContent-Type: %s",
|
|
file->contenttype);
|
|
}
|
|
if(file->contenttype &&
|
|
!strnequal("text/", file->contenttype, 5)) {
|
|
/* this is not a text content, mention our binary encoding */
|
|
size += AddFormData(&form, "\r\nContent-Transfer-Encoding: binary", 0);
|
|
}
|
|
|
|
|
|
size += AddFormData(&form, "\r\n\r\n", 0);
|
|
|
|
if((post->flags & HTTPPOST_FILENAME) ||
|
|
(post->flags & HTTPPOST_READFILE)) {
|
|
/* we should include the contents from the specified file */
|
|
FILE *fileread;
|
|
char buffer[1024];
|
|
int nread;
|
|
|
|
fileread = strequal("-", file->contents)?stdin:
|
|
/* binary read for win32 crap */
|
|
fopen(file->contents, "rb");
|
|
if(fileread) {
|
|
while((nread = fread(buffer, 1, 1024, fileread))) {
|
|
size += AddFormData(&form,
|
|
buffer,
|
|
nread);
|
|
}
|
|
if(fileread != stdin)
|
|
fclose(fileread);
|
|
} else {
|
|
size += AddFormData(&form, "[File wasn't found by client]", 0);
|
|
}
|
|
} else {
|
|
/* include the contents we got */
|
|
size += AddFormData(&form, post->contents, 0);
|
|
}
|
|
} while((file = file->more)); /* for each specified file for this field */
|
|
|
|
if(post->more) {
|
|
/* this was a multiple-file inclusion, make a termination file
|
|
boundary: */
|
|
size += AddFormDataf(&form,
|
|
"\r\n--%s--",
|
|
fileboundary);
|
|
free(fileboundary);
|
|
}
|
|
|
|
} while((post=post->next)); /* for each field */
|
|
|
|
/* end-boundary for everything */
|
|
size += AddFormDataf(&form,
|
|
"\r\n--%s--\r\n",
|
|
boundary);
|
|
|
|
*sizep = size;
|
|
|
|
free(boundary);
|
|
|
|
return firstform;
|
|
}
|
|
|
|
int FormInit(struct Form *form, struct FormData *formdata )
|
|
{
|
|
if(!formdata)
|
|
return 1; /* error */
|
|
|
|
#if 0
|
|
struct FormData *lastnode=formdata;
|
|
|
|
/* find the last node in the list */
|
|
while(lastnode->next) {
|
|
lastnode = lastnode->next;
|
|
}
|
|
|
|
/* Now, make sure that we'll send a nice terminating sequence at the end
|
|
* of the post. We *DONT* add this string to the size of the data since this
|
|
* is actually AFTER the data. */
|
|
AddFormDataf(&lastnode, "\r\n\r\n");
|
|
#endif
|
|
form->data = formdata;
|
|
form->sent = 0;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* fread() emulation */
|
|
int FormReader(char *buffer,
|
|
size_t size,
|
|
size_t nitems,
|
|
FILE *mydata)
|
|
{
|
|
struct Form *form;
|
|
int wantedsize;
|
|
int gotsize;
|
|
|
|
form=(struct Form *)mydata;
|
|
|
|
wantedsize = size * nitems;
|
|
|
|
if(!form->data)
|
|
return -1; /* nothing, error, empty */
|
|
|
|
do {
|
|
|
|
if( (form->data->length - form->sent ) > wantedsize ) {
|
|
|
|
memcpy(buffer, form->data->line + form->sent, wantedsize);
|
|
|
|
form->sent += wantedsize;
|
|
|
|
return wantedsize;
|
|
}
|
|
|
|
memcpy(buffer,
|
|
form->data->line + form->sent,
|
|
gotsize = (form->data->length - form->sent) );
|
|
|
|
form->sent = 0;
|
|
|
|
form->data = form->data->next; /* advance */
|
|
|
|
} while(!gotsize && form->data);
|
|
/* If we got an empty line and we have more data, we proceed to the next
|
|
line immediately to avoid returning zero before we've reached the end.
|
|
This is the bug reported November 22 1999 on curl 6.3. (Daniel) */
|
|
|
|
return gotsize;
|
|
}
|
|
|
|
|
|
#ifdef _FORM_DEBUG
|
|
|
|
int main(int argc, char **argv)
|
|
{
|
|
#if 0
|
|
char *testargs[]={
|
|
"name1 = data in number one",
|
|
"name2 = number two data",
|
|
"test = @upload"
|
|
};
|
|
#endif
|
|
int i;
|
|
char *nextarg;
|
|
struct HttpPost *httppost=NULL;
|
|
struct HttpPost *last_post=NULL;
|
|
struct HttpPost *post;
|
|
int size;
|
|
int nread;
|
|
char buffer[4096];
|
|
|
|
struct FormData *form;
|
|
struct Form formread;
|
|
|
|
for(i=1; i<argc; i++) {
|
|
|
|
if( FormParse( argv[i],
|
|
&httppost,
|
|
&last_post)) {
|
|
fprintf(stderr, "Illegally formatted input field: '%s'!\n",
|
|
argv[i]);
|
|
return 1;
|
|
}
|
|
}
|
|
|
|
form=getFormData(httppost, &size);
|
|
|
|
FormInit(&formread, form);
|
|
|
|
while(nread = FormReader(buffer, 1, sizeof(buffer), (FILE *)&formread)) {
|
|
fwrite(buffer, nread, 1, stderr);
|
|
}
|
|
|
|
fprintf(stderr, "size: %d\n", size);
|
|
|
|
return 0;
|
|
}
|
|
|
|
#endif
|