cpp: malformed key names not rejected

[markus2330]

Steps to Reproduce the Problem

import kdb
k = kdb.Key("abc")
k.name="abc"

Expected Result

That for both ways to assign the key's name an exception is raised that "abc" is not a valid keyname.

Actual Result

k = kdb.Key("abc")
#> <kdb.Key; proxy of <Swig Object of type 'kdb::Key *' at 0x7f9ec38b2c90> >
k.name
#> ''
k.name="abc"
#> segfault

System Information

• Elektra Version: master

[manuelm]

I've modified the title as this is no python nor swig issue. A possible fix, feel free to commit this:

--- a/src/bindings/cpp/include/key.hpp
+++ b/src/bindings/cpp/include/key.hpp
@@ -604,6 +604,7 @@ inline Key::Key (const char * keyName, ...)
        va_end (ap);
 
        if (!key) throw std::bad_alloc ();
+       if (ckdb::keyGetUnescapedNameSize (key) <= 0) throw KeyInvalidName ();
 
        operator++ ();
 }
@@ -628,6 +629,7 @@ inline Key::Key (const std::string keyName, ...)
        va_end (ap);
 
        if (!key) throw std::bad_alloc ();
+       if (ckdb::keyGetUnescapedNameSize (key) <= 0) throw KeyInvalidName ();
 
        operator++ ();
 }
@@ -645,6 +647,7 @@ inline Key::Key (const char * keyName, va_list ap)
        key = ckdb::keyVNew (keyName, ap);
 
        if (!key) throw std::bad_alloc ();
+       if (ckdb::keyGetUnescapedNameSize (key) <= 0) throw KeyInvalidName ();
 
        operator++ ();
 }

[markus2330]

Thank you! Can you please add this commit to your PR?

[manuelm]

@markus2330 As expected changing the ctor to throw on invalid names makes tests fail on other places. Can you or someone else please take over ea8a01e. I feel like I've already invested enough time in this.

[markus2330]

Actually this problem will solve itself once #3115 is merged. Because then keyNew will fail with malformed key names.