{device,brand,module}_replacement can refer to non-existent capture groups

[jimdigriz]

We found '; *(Google )?(Nexus [Ss](?: 4G)?) Build/' failed as both device_replacement and model_replacement can refer to none existent capture groups (if Google is not present then $2 is not defined).

This probably should be rewritten (untested and off the top of my head) as ; *(Google |)(Nexus [Ss](?: 4G)?) Build/ to avoid this.

Really crudely, looks like someone will need to flip through additionally the following to check at least these are safe, but someone is going to have to come up with a regex to test your regex's...Dawg :) Should help in future as they get updated and added to so problems like this do not crop up again:

$ grep -A3 '([^(?:)][^()]*)?' regexes.yaml | grep -B3 '_replacement.*\$' | grep regex:
  - regex: '(Namoroka|Shiretoko|Minefield)/(\d+)\.(\d+)([ab]\d+[a-z]*)?'
  - regex: 'Android Application[^\-]+ - (Sony) ?(Ericsson)? (.+) \w+ - '
  - regex: '; *(Advent )?(Vega(?:Bean|Comb)?).* Build'
  - regex: '; *(Ainol )?((?:NOVO|[Nn]ovo)[^;/]+) Build'
  - regex: '; *(ALLVIEW[ _]?|Allview[ _]?)?(AX1_Shine|AX2_Frenzy) Build'
  - regex: '; *(CUBE[ _])?([KU][0-9]+ ?GT.*|A5300) Build'
  - regex: '; *(HUAWEI |Huawei-)?([UY][^;/]+) Build/(?:Huawei|HUAWEI)([UY][^\);]+)\)'
  - regex: '; *(MODECOM )?(FreeTab) ?([^;/]+) Build'
  - regex: '; *(A\d+)[ _](Duo)? Build'
  - regex: '; *(NOOK )?(BNRV200|BNRV200A|BNTV250|BNTV250A|BNTV400|BNTV600|LogicPD Zoom2) Build'
  - regex: '; *(SKY[ _])?(IM\-[AT]\d{3}[^;/]+).* Build/'
  - regex: 'Android 4\..*; *(M[12356789]|U[12368]|S[123])\ ?(pro)? Build'
  - regex: '; *(?:Polaroid[ _])?((?:MIDC\d{3,}|PMID\d{2,}|PTAB\d{3,})[^;/]*)(\/[^;/]*)? Build/'
  - regex: '; *(A2|A5|A8|A900)_?(Classic)? Build'
  - regex: '; *(SAMSUNG |Samsung )?((?:Galaxy (?:Note II|S\d)|GT-I9082|GT-I9205|GT-N7\d{3}|SM-N9005)[^;/]*)\/?[^;/]* Build/'
  - regex: '; *(Google )?(Nexus [Ss](?: 4G)?) Build/'
  - regex: '; *(SAMSUNG-)?(GT\-[BINPS]\d{4}[^\/]*)(\/[^ ]*) Build'
  - regex: '; *((?:SCH|SGH|SHV|SHW|SPH|SC|SM)\-[A-Za-z0-9 ]+)(/?[^ ]*)? Build'
  - regex: ' ((?:SCH)\-[A-Za-z0-9 ]+)(/?[^ ]*)? Build'
  - regex: '; *((?:CSL_Spice|Spice|SPICE|CSL)[ _\-]?)?([Mm][Ii])([ _\-])?(\d{3}[^;/]*) Build/'
  #~ - regex: '; *(Sprint)? ?(HTC_?)?(X515E|ATP515CKIT|APA7373KT|PG06100|APC715CKT|APX515CKT|PG86100|EVOV4G|APA9292KT|PC36100) Build'
  - regex: '\b(T-Mobile ?)?(myTouch)[ _]?([34]G)[ _]?([^\/]*) (?:Mozilla|Build)'

[commenthol]

Could need a hint on which parser you encountered the failures. Using uap-ref-impl this does not seam to be an issue.

[jimdigriz]

The python user_agent_parser.

As a passing question, in the case of the following regex matching some string (say moo cow):

([^ ]+)( [^ ]+)? ([^ ]+)

What would you say $3 should be?

IIRC both Perl and Python will only define $3 if the second capture matches something, otherwise you get undefined variables flying around.

[jimdigriz]

Confusion from mis-reading specification