…ots are given

The API specifically documents that the caller may provide *any* number
of slots. But the onepass implementation was panicking when too many
were provided. (I had only considered the case when too few were
provided when I wrote one-pass implementation I think.)

This already works for the PikeVM and the bounded backtracker. This
commit adds tests covering those as well.

Note that the bug report specifically cites a zero-repetition capture
group as the instigator here. While that may have provided the
circumstances for the bug to appear, the actual bug is more general than
that. You don't need any capture groups at all to provoke it. You just
need to provide more slots than there are in the compiled regex.

Fixes #1327, Closes #1329

…ots are given

The API specifically documents that the caller may provide *any* number
of slots. But the onepass implementation was panicking when too many
were provided. (I had only considered the case when too few were
provided when I wrote one-pass implementation I think.)

This already works for the PikeVM and the bounded backtracker. This
commit adds tests covering those as well.

Note that the bug report specifically cites a zero-repetition capture
group as the instigator here. While that may have provided the
circumstances for the bug to appear, the actual bug is more general than
that. You don't need any capture groups at all to provoke it. You just
need to provide more slots than there are in the compiled regex.

Fixes #1327, Closes #1329

…ots are given

The API specifically documents that the caller may provide *any* number
of slots. But the onepass implementation was panicking when too many
were provided. (I had only considered the case when too few were
provided when I wrote one-pass implementation I think.)

This already works for the PikeVM and the bounded backtracker. This
commit adds tests covering those as well.

Note that the bug report specifically cites a zero-repetition capture
group as the instigator here. While that may have provided the
circumstances for the bug to appear, the actual bug is more general than
that. You don't need any capture groups at all to provoke it. You just
need to provide more slots than there are in the compiled regex.

Fixes #1327, Closes #1329

…ots are given

The API specifically documents that the caller may provide *any* number
of slots. But the onepass implementation was panicking when too many
were provided. (I had only considered the case when too few were
provided when I wrote one-pass implementation I think.)

This already works for the PikeVM and the bounded backtracker. This
commit adds tests covering those as well.

Note that the bug report specifically cites a zero-repetition capture
group as the instigator here. While that may have provided the
circumstances for the bug to appear, the actual bug is more general than
that. You don't need any capture groups at all to provoke it. You just
need to provide more slots than there are in the compiled regex.

Fixes #1327, Closes #1329

…ots are given

The API specifically documents that the caller may provide *any* number
of slots. But the onepass implementation was panicking when too many
were provided. (I had only considered the case when too few were
provided when I wrote one-pass implementation I think.)

This already works for the PikeVM and the bounded backtracker. This
commit adds tests covering those as well.

Note that the bug report specifically cites a zero-repetition capture
group as the instigator here. While that may have provided the
circumstances for the bug to appear, the actual bug is more general than
that. You don't need any capture groups at all to provoke it. You just
need to provide more slots than there are in the compiled regex.

Fixes rust-lang#1327, Closes rust-lang#1329