Skip to contentSkip to conversation

Build system improvement

PATCH In review Javier Martín 9 participants 16 messages Jan 25, 2009 – Jul 22, 2009 Kbuild

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Build system improvement

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

Makefile.in +16 -4
Index: Makefile.in
===================================================================
@@ -138,18 +138,30 @@
 CLEANFILES += $(pkglib_DATA) $(pkgdata_DATA)
 pkglib_DATA += moddep.lst command.lst fs.lst partmap.lst
 moddep.lst: $(DEFSYMFILES) $(UNDSYMFILES) genmoddep.awk
-	cat $(DEFSYMFILES) /dev/null \
+ifneq ($(V),1)
+	@echo [L moddep.lst] Building MODULE DEPENDENCIES list
+endif
+	$(SILENT)cat $(DEFSYMFILES) /dev/null \
 	  | $(AWK) -f $(srcdir)/genmoddep.awk $(UNDSYMFILES) > $@ \
 	  || (rm -f $@; exit 1)
 
 command.lst: $(COMMANDFILES)
-	cat $^ /dev/null | sort > $@
+ifneq ($(V),1)
+	@echo [L command.lst] Building COMMAND list
+endif
+	$(SILENT)cat $^ /dev/null | sort > $@
 
 fs.lst: $(FSFILES)
-	cat $^ /dev/null | sort > $@
+ifneq ($(V),1)
+	@echo [L fs.lst] Building FILESYSTEM list
+endif
+	$(SILENT)cat $^ /dev/null | sort > $@
 
 partmap.lst: $(PARTMAPFILES)
-	cat $^ /dev/null | sort > $@
+ifneq ($(V),1)
+	@echo [L partmap.lst] Building PARTMAP list
+endif
+	$(SILENT)cat $^ /dev/null | sort > $@
 
 ifeq (, $(UNIFONT_BDF))
 else
common.rmk +13 -8
Index: conf/common.rmk
===================================================================
@@ -1,5 +1,9 @@
 # -*- makefile -*-
 
+ifneq ($(V),1)
+SILENT = @
+endif
+
 # For grub-mkelfimage.
 bin_UTILITIES += grub-mkelfimage
 grub_mkelfimage_SOURCES = util/elf/grub-mkimage.c util/misc.c \
@@ -125,26 +129,27 @@
 
 # For grub-mkconfig
 grub-mkconfig: util/grub-mkconfig.in config.status
-	./config.status --file=$@:$<
-	chmod +x $@
+	$(SILENT)./config.status --file=$@:$<
+	$(SILENT)chmod +x $@
 sbin_SCRIPTS += grub-mkconfig
 CLEANFILES += grub-mkconfig
 
 grub-mkconfig_lib: util/grub-mkconfig_lib.in config.status
-	./config.status --file=$@:$<
-	chmod +x $@
+	$(SILENT)./config.status --file=$@:$<
+	$(SILENT)chmod +x $@
 lib_DATA += grub-mkconfig_lib
 CLEANFILES += grub-mkconfig_lib
 
 update-grub_lib: util/update-grub_lib.in config.status
-	./config.status --file=$@:$<
-	chmod +x $@
+	$(SILENT)./config.status --file=$@:$<
+	$(SILENT)chmod +x $@
 lib_DATA += update-grub_lib
 CLEANFILES += update-grub_lib
 
 %: util/grub.d/%.in config.status
-	./config.status --file=$@:$<
-	chmod +x $@
+#no need to echo nothing here: config.status already does so
+	$(SILENT)./config.status --file=$@:$<
+	$(SILENT)chmod +x $@
 grub-mkconfig_SCRIPTS = 00_header 10_linux 10_hurd 10_freebsd 30_os-prober 40_custom
 ifeq ($(target_os), cygwin)
 grub-mkconfig_SCRIPTS += 10_windows
genmk.rb +75 -23

181 lines not on this page · raw message

#1 lore raw
Quoted 18 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

Quoted 3 lines

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

First of all I would like compiling process to be more tidier.

However, what is the difference between Colin's similar patch?

#2 lore raw
Quoted 19 lines · Vesa Jääskeläinen
Quoted 18 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

Quoted 7 lines
Quoted 3 lines

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

First of all I would like compiling process to be more tidier.

However, what is the difference between Colin's similar patch?

Oops... I did not remember Colin's patch, it's been so long since it was last discussed that my mind considered it deceased for all intents and purposes. I'll look into the post history and see what it did compared to mine - if I get it to compile against a current SVN grub, that is.

#3 lore raw
Quoted 24 lines · On Sun, 25 Jan 2009 14:23:29 +0100 Javier Martín <[email protected]> wrote:
Quoted 23 lines · Vesa Jääskeläinen
Quoted 21 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

Quoted 7 lines
Quoted 3 lines

First of all I would like compiling process to be more tidier.

However, what is the difference between Colin's similar patch?

Oops... I did not remember Colin's patch, it's been so long since it was last discussed that my mind considered it deceased for all intents and purposes. I'll look into the post history and see what it did compared to mine - if I get it to compile against a current SVN grub, that is.

Basically the difference is that my 'prettymake' patch redefines the program variables like TARGET_CC, etc. so that they cause the echo of a nice like when they are executed.

This is the main part of the patch (from Makefile.in)

+### Pretty output control ###
+# Set up compiler and linker commands that either is quiet (does not print +# the command line being executed) or verbose (print the command line). +_CC := $(CC)
+_TARGET_CC := $(TARGET_CC)
+_STRIP := $(STRIP)
+_GENMODSRC := sh $(srcdir)/genmodsrc.sh
+ifeq ($(V),1)
+ override V_PREFIX :=
+ override CC = $(_CC)
+ override TARGET_CC = $(_CC)
+ override STRIP = $(_STRIP)
+ override GENMODSRC = $(_GENMODSRC)
+ override INFO_GENCMDLIST =
+ override INFO_GENFSLIST =
+ override INFO_GENPARTMAPLIST =
+ override INFO_GEN_FINAL_COMMAND_LIST =
+ override INFO_GEN_FINAL_FS_LIST =
+ override INFO_GEN_FINAL_PARTMAP_LIST =
+else
+ override V_PREFIX := @
+ override CC = @echo "COMPILE $<"; $(_CC)
+ override TARGET_CC = @echo "COMPILE(TARGET) $<"; $(_TARGET_CC)
+ override STRIP = @echo "STRIP $@"; $(_STRIP)
+ override GENMODSRC = @echo "GENMODSRC $@"; $(_GENMODSRC)
+ override INFO_GENCMDLIST = @echo "GENCMDLIST $@"
+ override INFO_GENFSLIST = @echo "GENFSLIST $@"
+ override INFO_GENPARTMAPLIST = @echo "GENPARTMAPLIST $@"
+ override INFO_GEN_FINAL_COMMAND_LIST = @echo "GENCMDLIST[final] $@"
+ override INFO_GEN_FINAL_FS_LIST = @echo "GENFSLIST[final] $@"
+ override INFO_GEN_FINAL_PARTMAP_LIST = @echo "GENPARTMAPLIST[final] $@" +endif

I implemented it this way because it was easiest, since I didn't have to hunt down all the usages of the tools to add the output control.

This method has downsides, and I think everyone will agree Javier's method is much more robust and more transparent as to what it is doing. Also, my patch no longer works, which shows its fragility: For instance, in current GRUB SVN the value of TARGET_CC is being referenced outside the context of actually executing it (its value is being assigned to another variable). This breaks with my patch because of the way TARGET_CC is redefined.

If you want to see my whole patch (against older GRUB SVN -- it doesn't quite work against current GRUB for the reason mentioned above), you can see it at
<http://grub.gibibit.com/patches/prettymake_2008-07-04.patch>

Regards,
Colin

#4 lore raw
In reply to Javier Martín

I don't see any stoppers to merge this patch

Quoted 27 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c ->
xfs_mod-fs_xfs.o [M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o [M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

------------------------------------------------------------------------

_______________________________________________
Grub-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/grub-devel

Signature · 2 lines
Regards
Vladimir 'phcoder' Serbinenko
#5 lore raw
Quoted 1 line · On Saturday 11 April 2009 08:25:50 phcoder wrote:

I don't see any stoppers to merge this patch

Personally, I don't like this so much. Is it so useful?

Regards,
Okuji

Quoted 31 lines
Quoted 29 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

------------------------------------------------------------------------

_______________________________________________
Grub-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/grub-devel

#6 lore raw

The problem is that build system creates a lot of output on progress and so the warnings are often not very visible

Quoted 46 lines · Yoshinori K. Okuji
Quoted 1 line · On Saturday 11 April 2009 08:25:50 phcoder wrote:

I don't see any stoppers to merge this patch

Personally, I don't like this so much. Is it so useful?

Regards,
Okuji

Quoted 24 lines
Quoted 23 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

Quoted 6 lines
Quoted 6 lines

------------------------------------------------------------------------

_______________________________________________
Grub-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/grub-devel

_______________________________________________
Grub-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/grub-devel

Signature · 2 lines
Regards
Vladimir 'phcoder' Serbinenko
#7 lore raw
In reply to Yoshinori K. Okuji

Yoshinori K. Okuji wrote on Saturday 11 April 2009:

Quoted 4 lines · On Saturday 11 April 2009 08:25:50 phcoder wrote:
Quoted 1 line

I don't see any stoppers to merge this patch

Personally, I don't like this so much. Is it so useful?

Absolutely. I find it tedious to scroll back through thousands of lines of text full of gcc command lines with lots of options etc. to find the actual warnings that could indicate (and usually do!) that I made an error in my code. When all the lines generally have a concise and uniform appearance, it is visually easy to scan for warnings, compared to now where the lines printed are very long and often wrap onto multiple lines, obscuring warnings.

If we could build with -Werror, then it wouldn't be so hard to find the warnings since the build would abort...

I think there is only one warning (the “trampoline” warning) that we want to ignore. All others can and should be fixed (and they may be already).

Regards,
Colin

Quoted 24 lines
Quoted 24 lines
Quoted 23 lines · Javier Martín

This patch modifies several files in the build system (mainly common.rmk and genmk.rb) to reduce the general verbosity of the build process to a manageable, semi-informative level. Thus, what currently appears as "gcc" calls, several lines long each is turned into lines like:

[M xfs.mod] COMPILE ../src/fs/xfs.c -> xfs_mod-fs_xfs.o
[M xfs.mod] LINK xfs_mod-fs_xfs.o ->
pre-xfs.o
[M xfs.mod] Looking for EXPORTED SYMBOL definitions: pre-xfs.o

And so on. The change also makes warning-hunting marginally easier, though not by much since the patch intentionally shows a line for nearly every process that did so previously. This behavior could be simplified further if needed - this post is more of an RFC than anything else. Also, it is by no means thorough or complete - only the most common processes have been addressed - as I'm a bit busy with exams.

The patch makes the new behavior the default one, so a new make-time option is added: V (for "verbose"), which must have the value 1 in order to get the behavior, as in "make V=1"

#8 lore raw
Quoted 2 lines · Colin D Bennett

If we could build with -Werror, then it wouldn't be so hard to find the warnings since the build would abort...

It's also possible to redirect stderr to a file so that the build doesn't stumble on the first warning.

I my opinion, making the build less verbose will increase the probability that the warnings will be caught and fixed by their makers rather than by dedicated warning hunters who may not know the code so well as those who wrote it. And that's a good thing.

Signature · 2 lines
Regards,
Pavel Roskin
#9 lore raw

From: Pavel Roskin <[email protected]>
Date: Sun, 12 Apr 2009 17:24:49 -0400

Quoted 7 lines · Colin D Bennett
Quoted 3 lines

If we could build with -Werror, then it wouldn't be so hard to find the warnings since the build would abort...

It's also possible to redirect stderr to a file so that the build doesn't stumble on the first warning.

I'm iffy about this.

There are some hard warnings to get rid of.

For example when building certain grub-* tools there is no way to get around the current redefinitions we get of LONG_MAX and friends. (one comes in via grub headers, then the stdio.h include gets us the system definition, we can't use ifdef guards because the grub headers come in and define things first)

#10 lore raw
Quoted 5 lines · David Miller

From: Pavel Roskin <[email protected]>
Date: Sun, 12 Apr 2009 17:24:49 -0400

Quoted 2 lines · Colin D Bennett
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

If we could build with -Werror, then it wouldn't be so hard to find the warnings since the build would abort...

It's also possible to redirect stderr to a file so that the build doesn't stumble on the first warning.

I'm iffy about this.

I meant that "warning hunters" can use it and have a choice what warnings to fix. I didn't suggest stderr redirection to be part of the build system.

Quoted 7 lines

There are some hard warnings to get rid of.

For example when building certain grub-* tools there is no way to get around the current redefinitions we get of LONG_MAX and friends. (one comes in via grub headers, then the stdio.h include gets us the system definition, we can't use ifdef guards because the grub headers come in and define things first)

I would explore the possibility of introducing GRUB_LONG_MAX. GRUB already duplicates a lot of libc definitions.

Sure enough, forcing developers to keep the code warning-free could lead to ugly fixes that are worse that the original warnings. However, keeping the build system quiet will help more people notice the warnings. Somebody will come with a nice fix eventually.

Signature · 2 lines
Regards,
Pavel Roskin
#11 lore raw
Quoted 19 lines · On Monday 13 April 2009 14:03:01 Pavel Roskin wrote:
Quoted 11 lines · David Miller

From: Pavel Roskin <[email protected]>
Date: Sun, 12 Apr 2009 17:24:49 -0400

Quoted 6 lines · Colin D Bennett
Quoted 2 lines

If we could build with -Werror, then it wouldn't be so hard to find the warnings since the build would abort...

It's also possible to redirect stderr to a file so that the build doesn't stumble on the first warning.

I'm iffy about this.

I meant that "warning hunters" can use it and have a choice what warnings to fix. I didn't suggest stderr redirection to be part of the build system.

Quoted 2 lines

There are some hard warnings to get rid of.

Quoted 8 lines
Quoted 5 lines

For example when building certain grub-* tools there is no way to get around the current redefinitions we get of LONG_MAX and friends. (one comes in via grub headers, then the stdio.h include gets us the system definition, we can't use ifdef guards because the grub headers come in and define things first)

I would explore the possibility of introducing GRUB_LONG_MAX. GRUB already duplicates a lot of libc definitions.

Yes. It is bad and dangerous to use the same symbols as libc. I think I have written this in the wiki:

http://grub.enbug.org/CodingStyle

Regards,
Okuji

#12 lore raw
In reply to Yoshinori K. Okuji

Am Samstag, den 11.04.2009, 19:09 +0900 schrieb Yoshinori K. Okuji:

Quoted 5 lines · On Saturday 11 April 2009 08:25:50 phcoder wrote:
Quoted 1 line

I don't see any stoppers to merge this patch

Personally, I don't like this so much. Is it so useful?

Okuji can we please have this?
Robert brought up again to use -Werror in the `Code quality' topic [0] and as a start it would be very useful to make the warnings more visible by default.

[0] http://lists.gnu.org/archive/html/grub-devel/2009-05/msg00009.html

Signature · 1 line
Felix Zielcke
#13 lore raw

Am Sonntag, den 03.05.2009, 10:29 +0200 schrieb Felix Zielcke:

Quoted 7 lines

Am Samstag, den 11.04.2009, 19:09 +0900 schrieb Yoshinori K. Okuji:

Quoted 5 lines · On Saturday 11 April 2009 08:25:50 phcoder wrote:
Quoted 1 line

I don't see any stoppers to merge this patch

Personally, I don't like this so much. Is it so useful?

Quoted 6 lines

Okuji can we please have this?
Robert brought up again to use -Werror in the `Code quality' topic [0] and as a start it would be very useful to make the warnings more visible by default.

[0] http://lists.gnu.org/archive/html/grub-devel/2009-05/msg00009.html

So what's now with it?
Marco can you at least say something about it?
It would be very nice if the warnings would be more visible by default. Probable we can't ever use -Werror because different compilers produce different warnings etc.
For example gentoo's gcc with his trampoline patch.

Signature · 1 line
Felix Zielcke
#14 lore raw
Quoted 8 lines · Felix Zielcke
Quoted 1 line

[0] http://lists.gnu.org/archive/html/grub-devel/2009-05/msg00009.html

So what's now with it?
Marco can you at least say something about it?
It would be very nice if the warnings would be more visible by default. Probable we can't ever use -Werror because different compilers produce different warnings etc.
For example gentoo's gcc with his trampoline patch.

I don't see any warnings in any configuration, unless more warning flags are added (e.g. -Wconversion, which is very noisy).

We can use -Werror by default, but make it very easy to disable. For instance, we could have a configure option --disable-strict-warnings.

GRUB is a bootloader; we don't want anyone's system to become unbootable because something was miscompiled. Requiring an extra step to disable warnings would tell users to be more careful.

Signature · 2 lines
Regards,
Pavel Roskin
#15 lore raw
Quoted 14 lines · Pavel Roskin
Quoted 8 lines · Felix Zielcke
Quoted 1 line

[0] http://lists.gnu.org/archive/html/grub-devel/2009-05/msg00009.html

So what's now with it?
Marco can you at least say something about it?
It would be very nice if the warnings would be more visible by default. Probable we can't ever use -Werror because different compilers produce different warnings etc.
For example gentoo's gcc with his trampoline patch.

I don't see any warnings in any configuration, unless more warning flags are added (e.g. -Wconversion, which is very noisy).

Quoted 6 lines

We can use -Werror by default, but make it very easy to disable. For instance, we could have a configure option --disable-strict-warnings.

GRUB is a bootloader; we don't want anyone's system to become unbootable because something was miscompiled. Requiring an extra step to disable warnings would tell users to be more careful.

I completely agree with this (both things).

Except, I'd call it --disable-werror, I think it's clearer (and I vaguely recall other projects using this flag).

Signature · 5 lines
Robert Millan

  The DRM opt-in fallacy: "Your data belongs to us. We will decide when (and
  how) you may access your data; but nobody's threatening your freedom: we
  still allow you to remove your data and not access it at all."