REGRESSION(r46060): Error generating CSSValueKeywords.h during WebCoreGenerated

[Mark Rowe (bdash)]

Created attachment 33121 [details]
Full transcript from WebCoreGenerated project

After <http://trac.webkit.org/changeset/46060> a number of people at Apple are seeing an error during WebCoreGenerated that causes WebCore to fail to build.  Some relevant snippets from the build log are as follows (a more complete log is attached):

16>Creating hashtable for /home/Jessie/OPENSO~1/WebCore/bindings/js/JSRGBColor.cpp
16>bison -d -p cssyy /home/Jessie/OPENSO~1/WebCore/css/CSSGrammar.y -o CSSGrammar.cpp
16>if sort /home/Jessie/OPENSO~1/WebCore/css/CSSPropertyNames.in /home/Jessie/OPENSO~1/WebCore/css/SVGCSSPropertyNames.in | uniq -d | grep -E '^[^#]'; then echo 'Duplicate value!'; exit 1; fi
16>touch CSSGrammar.cpp.h
16>touch CSSGrammar.hpp
16>echo '#ifndef CSSGrammar_h' > CSSGrammar.h
16>Invalid switch.
16>cat /home/Jessie/OPENSO~1/WebCore/css/CSSPropertyNames.in /home/Jessie/OPENSO~1/WebCore/css/SVGCSSPropertyNames.in > CSSPropertyNames.in
16>echo '#define CSSGrammar_h' >> CSSGrammar.h
16>cat CSSGrammar.cpp.h CSSGrammar.hpp >> CSSGrammar.h
16>perl "/home/Jessie/OPENSO~1/WebCore/css/makeprop.pl"
16>echo '#endif' >> CSSGrammar.h
16>rm -f CSSGrammar.cpp.h CSSGrammar.hpp
16># Lower case all the values, as CSS values are case-insensitive
16>gperf -CDEot -L ANSI-C -k '*' -N findColor -D -s 2 /home/Jessie/OPENSO~1/WebCore/platform/ColorData.gperf > ColorData.c
16>perl -ne 'print lc' /home/Jessie/OPENSO~1/WebCore/css/CSSValueKeywords.in /home/Jessie/OPENSO~1/WebCore/css/SVGCSSValueKeywords.in > CSSValueKeywords.in
16>if sort CSSValueKeywords.in | uniq -d | grep -E '^[^#]'; then echo 'Duplicate value!'; exit 1; fi
16>gperf -CEot -L ANSI-C -k "*" -N findDoctypeEntry -F ,PubIDInfo::eAlmostStandards,PubIDInfo::eAlmostStandards /home/Jessie/OPENSO~1/WebCore/html/DocTypeStrings.gperf > DocTypeStrings.cpp
16>Duplicate value!
16>make: *** [CSSValueKeywords.h] Error 1
16>make: *** Waiting for unfinished jobs….

A few interesting lines jump out:
16>Invalid switch.
16>Duplicate value!

This doesn't reproduce on the build.webkit.org build bot in a clean build so it may be in some way related to the differences between the Debug_Internal and Debug configurations.

[Peter Kasting]

Backed out r46060 in r46139.

So far no one seems to know precisely how r46060 causes this, so if someone who was having the problem (mitz? JessieBerlin?) has time to work with me privately, it'd be nice to find a way to reland that.

[David Kilzer (:ddkilzer)]

http://trac.webkit.org/changeset/46060