Skip to contentSkip to conversation

igt: Fix detection of missing flex

PATCH In review Tvrtko Ursulin 2 participants 2 messages May 26, 2017 – May 30, 2017

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 igt: Fix detection of missing flex

From: Tvrtko Ursulin <[email protected]>

AM_PROG_FLEX macro will set the LEX variable using the missing script when the flex is not present. This will confuse the configure.ac check, which expects the AC_PROG_FLEX behaviour, and will so fail to detect the missing flex:

	AS_IF([test x"$LEX" != "x:" -a x"$YACC" != xyacc],
		[enable_assembler=yes],
		[enable_assembler=no])

This is because AM_PROG_LEX sets the LEX variable to "${SHELL} /home/sc/intel-gpu-tools/build-aux/missing flex", while AC_PROG_LEX would set it to ":".

If for some reason we really need to keep AM_PROG_LEX, alternative fix could be something like this placed before the above AS_IF check:

	AC_MSG_CHECKING([checking for working flex])
	if ! eval "$LEX --version >/dev/null 2>&1"; then
		AC_MSG_RESULT([failed])
		LEX=:
	else
		AC_MSG_RESULT([pass])
	fi

Note the evil eval needed to recursively expand variables.

---
 configure.ac | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
configure.ac +1 -1
index 5342e33c8293..82857a8325b1 100644
@@ -40,7 +40,7 @@ AM_INIT_AUTOMAKE([1.12 foreign subdir-objects dist-bzip2])
 AM_PATH_PYTHON([3],, [:])
 
 AC_PROG_CC
-AM_PROG_LEX
+AC_PROG_LEX
 AC_PROG_YACC
 
 # check for gtk-doc
Signature · 6 lines
2.9.4

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 1/1
#1 lore raw
Quoted 7 lines · Tvrtko Ursulin

From: Tvrtko Ursulin <[email protected]>

AM_PROG_FLEX macro will set the LEX variable using the missing script when the flex is not present. This will confuse the configure.ac check, which expects the AC_PROG_FLEX behaviour, and will so fail to detect the missing flex:

Quoted 7 lines
	AS_IF([test x"$LEX" != "x:" -a x"$YACC" != xyacc],
		[enable_assembler=yes],
		[enable_assembler=no])

This is because AM_PROG_LEX sets the LEX variable to "${SHELL} /home/sc/intel-gpu-tools/build-aux/missing flex", while AC_PROG_LEX would set it to ":".

Good catch!

Quoted 13 lines

If for some reason we really need to keep AM_PROG_LEX, alternative fix could be something like this placed before the above AS_IF check:

	AC_MSG_CHECKING([checking for working flex])
	if ! eval "$LEX --version >/dev/null 2>&1"; then
		AC_MSG_RESULT([failed])
		LEX=:
	else
		AC_MSG_RESULT([pass])
	fi
Quoted 1 line

Note the evil eval needed to recursively expand variables.

As far as I can tell, A*_PROG_LEX is only used to be able to select whether to build the assembler. In other words, we don't need to AM_PROG_LEX behaviour.

There's still a bug left though: If flex is not found, but lex is, AC_PROG_LEX sets $LEX to 'lex', but assembler/Makefile.am hardcodes 'flex -i' anyway if the assembler is enabled. That still happens whether AM_ or AC_ is used so this patch is still an improvement.

I would also like to note that the ratio of characters-in-commit-message to characters-changed-in-code is quite the winner. :P

Quoted 26 lines
---
 configure.ac | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
configure.ac +1 -1
index 5342e33c8293..82857a8325b1 100644
@@ -40,7 +40,7 @@ AM_INIT_AUTOMAKE([1.12 foreign subdir-objects dist-bzip2])
 AM_PATH_PYTHON([3],, [:])
 
 AC_PROG_CC
-AM_PROG_LEX
+AC_PROG_LEX
 AC_PROG_YACC
 
 # check for gtk-doc
Signature · 6 lines
2.9.4

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx