In a precedence declaration, when there was a mix of litteral (e.g. 'a') and non-litteral (e.g. B) token names, Bison would number the litteral ones before the non-litteral ones, causing a numbering inversion in certain cases. For example, in the grammar: %right A B 'c' 'd' the tokens were numbered in this order: 'c' 'd' A B.
Valentin Tolmer (2): gram: correct token numbering in precedence declarations regen NEWS | 8 + src/parse-gram.c | 507 ++++++++++++++++++++++++++-------------------------- src/parse-gram.h | 14 +- src/parse-gram.y | 13 +- tests/conflicts.at | 86 +++++++++ 5 files changed, 369 insertions(+), 259 deletions(-) -- 1.7.9.5
