gschem segfaults on recursive loop

[KaiMartin]

This bug was reported on gmane.comp.cad.geda.user by Vladimir Zhbanov:

gschem segfaults when confronted with a symbol that contains itself.
To reproduce:
1) copy a symbol foobar.sym to the current working directory
2) open the copied symbol in gschem
3) type [i] and choose the local symbol foobar.sym in the chooser
4) add foobar.sym to the canvas
5) save (as foobar. sym) and close gschem
6) open gschem on foobar.sym
--> immediate segfault.

Suggestion: gschem should check for recursive loops when opening a schematic.

[Peter TB Brett (peter-b)]

Upgrading importance to "critical" because crashes are bad.

[Peter TB Brett (peter-b)]

Attached a testcase. To reproduce, run gschem broken.sym for an (almost) immediate crash.

[Peter TB Brett (peter-b)]

This bug needs to be fixed by loading schematic and symbol files in two passes. First, objects need to be parsed from the file while creating placeholder objects instead of complex objects, and *not* recursively loading symbols. Then, in a second pass, symbols need to be loaded from the library to convert the placeholders into "proper" symbols.

This approach would have three main benefits. Firstly, we can forbid symbols from containing symbols, fixing this bug. Secondly, applications would be able to more precisely report syntax errors etc. in symbols. Thirdly, it would provide an excuse to refactor and drastically simplify the libgeda file loading API.