Check libpng directly into third_party/

third_party/libpng/contrib/conftest/write.dfa

+# write.dfa
+#  Build time configuration of libpng
+#
+# Author: John Bowler
+# Copyright: (c) John Bowler, 2013
+# Usage rights:
+#  To the extent possible under law, the author has waived all copyright and
+#  related or neighboring rights to this work.  This work is published from:
+#  United States.
+#
+# Build libpng with no read support and minimal write support.
+#
+
+everything = off
+
+# Switch on the write code - this makes a minimalist encoder
+
+option WRITE on
+
+# Choose fixed or floating point APIs and arithmetic.  The choices are
+# independent but normally they will match.  It is typically better to use the
+# floating point if you have floating point hardware.  If you don't know, or
+# (perhaps) to make libpng smaller used fixed point throughout.
+
+#Fixed point:
+#option FIXED_POINT on
+#option FLOATING_ARITHMETIC off
+
+#Floating point:
+option FLOATING_POINT on
+option FLOATING_ARITHMETIC on
+
+# Basic error handling, IO and user memory support.  The latter allows the
+# application program to provide its own implementations of 'malloc' and 'free'.
+option SETJMP on
+option STDIO on
+option USER_MEM on
+
+# Everything else is optional.  Unlike the read code in libpng the write code
+# does not need to deal with arbitrary formats, so only add support for things
+# you really do write!  For example you might only write sRGB images, sometimes
+# with transparency and never write 16 bit images, so:
+option WRITE_sRGB on
+option WRITE_tRNS on
+#option WRITE_16BIT off (this is the default with 'everything = off')