oysteikt/heimdal
0
0
Fork 0
Code Issues 5 Pull Requests Releases Activity

asn1: Update commentary on grammar

Browse Source
This commit is contained in:
Nicolas Williams
2022-01-22 03:48:11 -06:00
parent dfc26ae8c1
commit d64076dfd9
1 changed files with 28 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
lib/asn1/asn1parse.y
View File

@@ -305,6 +305,9 @@ static unsigned long idcounter;
* We have sinned by allowing types to have names that start with lower-case,
* and values that have names that start with upper-case.
*
* UPDATE: We sin no more. However, parts of this block comment are still
* relevant.
*
* That worked when we only supported basic X.680 because the rules for
* TypeAssignment and ValueAssignment are clearly unambiguous in spite of the
* case issue.
@@ -313,35 +316,41 @@ static unsigned long idcounter;
* have to help us distinguish certain rules is the form of an identifier: the
* case of its first letter.
*
* We have begun to undo our sin by not allowing wrong-case identifiers in
* certain situations.
* We have cleansed our sin by not allowing wrong-case identifiers any more.
*
* Some historical instances of this sin in-tree:
*
* - DOMAIN-X500-COMPRESS (value (enum) but name starts with upper-case)
* - krb5int32 (type but name starts with lower-case)
* - krb5uint32 (type but name starts with lower-case)
* - hdb_keyset (type but name starts with lower-case)
* - hdb_entry (type but name starts with lower-case)
* - hdb_entry_alias (type but name starts with lower-case)
* - DOMAIN-X500-COMPRESS (value (enum) but name starts with upper-case)
* - krb5int32 (type but name starts with lower-case)
* - krb5uint32 (type but name starts with lower-case)
* - hdb_keyset (type but name starts with lower-case)
* - hdb_entry (type but name starts with lower-case)
* - hdb_entry_alias (type but name starts with lower-case)
* - HDB_DB_FORMAT INTEGER (value (int) but name starts with upper-case)
*
* We have fixed most of these, in some cases leaving behind aliases in header
* files as needed.
* We have fixed all of these and others, in some cases leaving behind aliases
* in header files as needed.
*
* This issue is probably also the source of remaining shift/reduce conflicts.
* We have one shift/reduce conflict (shift ObjectClassAssignment, reduce
* TypeAssignment) and one reduce/reduce conflict (ObjectAssignment vs
* ValueAssignment) that we avoid by requiring CLASS names to start with an
* underscore.
*
* In the FieldSetting rule in particular, we get a reduce/reduce conflict if
* we use `Identifier' instead of `TYPE_IDENTIFIER' for type field settings and
* In the FieldSetting rule, also, we get a reduce/reduce conflict if we use
* `Identifier' instead of `TYPE_IDENTIFIER' for type field settings and
* `VALUE_IDENTIFIER' for value field settings, and then we can't make
* progress.
*
* Looking forward, we may not (will not) be able to distinguish ValueSet and
* ObjectSet field settings from each other either even without committing this
* leading-identifier-character-case sin, and we may not (will not) be able
* distinguish Object and Value field settings from each other as well. To
* deal with those we will have to run-time type-tag/pun the C structures for
* valueset/objectset and value/object, and have one rule for each of those
* that inspects the type of the item to decide what kind of setting it is.
* ObjectSet field settings from each other either, and we may not (will not)
* be able distinguish Object and Value field settings from each other as well.
* To deal with those we will have to run-time type-tag and type-pun the C
* structures for valueset/objectset and value/object, and have one rule for
* each of those that inspects the type of the item to decide what kind of
* setting it is.
*
* Sadly, the extended syntax for ASN.1 (x.680 + x.681/2/3) appears to have
* ambiguities that cannot be resolved with bison/yacc.
*/
Identifier : TYPE_IDENTIFIER { $$ = $1; }
| VALUE_IDENTIFIER { $$ = $1; };