Arch Linux

This bug can be reproduced on ubuntu latest stardict package.
This problem can be fixed by deleting the youdaodict-plugin.
As upstream is not very active, I suggest skip installing the plugin in arch package as a temporary fix.

For segfaults, it's best to try and get a debug trace if possible. For this particular case, the plugin would need to be recompiled with debug symbols as the current trace is not much use (see attached).

I tried that and event did some live debuging. I believe I've found the root cause of the segfault.
In following function from dict/stardict-plugins/stardict-youdaodict-plugin/stardict_youdaodict.cpp

################################################################################
static void on_get_http_response(const char *buffer, size_t buffer_len, gpointer userdata)
{
if (!buffer) {
return;
}
const char *p = g_strstr_len(buffer, buffer_len, "\r\n\r\n");
if (!p) {
return;
}
p += 4;

QueryInfo *qi = (QueryInfo *)userdata;
NetDictResponse *resp = new NetDictResponse; <============================ Here, a new instance of NetDictResponse is created
resp->bookname = _("www.YouDao.com");
resp->booklink = "http://www.youdao.com";
resp->word = qi->word; // So neen't free qi->word;

process_xml_response(p, buffer_len - (p - buffer), resp); <================= The "data" field is supposed to be initialized in this function.

plugin_service->netdict_save_cache_resp(YOUDAODICT, qi->word, resp);
plugin_service->show_netdict_resp(YOUDAODICT, resp, qi->ismainwin);
delete qi;
keyword_list.remove(qi);
}
###############################################################################

But if the word is invalid, there won't be any search results. process_xml_response could return without setting the "data" field in resp. And no check against it.
In "void TextWin::Show(NetDictResponse *resp)" function from dict/src/mainwin.cpp, it actually checked the "data" pointer before using it.

###############################################################################
void TextWin::Show(NetDictResponse *resp)
{
view->begin_update();
bool do_append;
if (query_result == TEXT_WIN_FOUND || query_result == TEXT_WIN_SHOW_FIRST || query_result == TEXT_WIN_NET_FOUND || query_result == TEXT_WIN_NET_SHOW_FIRST) {
do_append = true;
view->goto_end();
} else {
do_append = false;
view->clear();
view->goto_begin();
}
if (resp->data) { <========================= Here, the "data" pointer is checked before use
query_result = TEXT_WIN_NET_FOUND;
InstantDictIndex dict_index;
dict_index.type = InstantDictType_UNKNOWN;
view->SetDictIndex(dict_index);
gchar *mark = g_strdup_printf("%d", view->get_bookindex());
gpAppFrame->oMidWin.oIndexWin.oResultWin.InsertLast(resp->bookname, mark);
g_free(mark);
view->AppendHeader(resp->bookname, resp->booklink);
view->AppendWord(resp->word);
view->AppendData(resp->data, resp->word, resp->word);
view->AppendNewline();
} else {
if (!do_append) {
query_result = TEXT_WIN_NET_NOT_FOUND;
Show(_("<Not Found!>"));
}
}
view->end_update();
}
#################################################################################
But as the "data" pointer is not initialized. It's random, not NULL. It will be used and trigger segfault.

So a possible fix is setting the data pointer to NULL in the first function I mentioned. I just tested the patch attached. No segfault anymore.