[LUCENE-8533] DataInput#readVInt() supports negative numbers although not documented

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 7.6, 8.0
• Lucene Fields: New, Patch Available
• Assignee: Uwe Schindler
• Reporter: Vladimir Dolzhenko
• Created: 16/Oct/18 10:05
• Updated: 28/Aug/22 15:36
• Resolved: 18/Oct/18 10:41

Description

readVInt() has to return positive numbers (and zero), throw some exception in case of negative numbers.

While for the sequence of bytes [-1, -1, -1, -1, 15] it returns -1.

simplifying readVInt up to last readByte (exclusive):

int i = ((byte)-1) & 0x7F;
i |= (((byte)-1) & 0x7F) << 7;
i |= (((byte)-1) & 0x7F) << 14;
i |= (((byte)-1) & 0x7F) << 21;

Here i = 268435455 or in binary format is 00001111_11111111_11111111_11111111

Keeping in mind that int is a signed type we have only 3 more bits before overflow happens or in another words (Integer.MAX_VALUE - i) >> 28 == 7 - that's max value could be stored in 5th byte to avoid overflow.

Instead of

i |= (b & 0x0F) << 28;
if ((b & 0xF0) == 0) return i;

has to be

i |= (b & 0x07) << 28;
if ((b & 0xF8) == 0) return i;