Skip to contentSkip to conversation

tiny-printf: Support %i

PATCH New Marek Vasut 3 participants 22 messages Apr 10, 2020 – May 1, 2020 Library code

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tiny-printf: Support %i

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +2 -1
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');
Signature · 1 line
2.25.1
#1 lore raw

PATCH 1/1 tiny-printf: Support %i

Hi Marek,

Quoted 10 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Can you add to the test in print_ut.c?

Regards,
Simon

#2 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 7 lines · Simon Glass

Hi Marek,

Quoted 4 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Quoted 8 lines
Quoted 6 lines
Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Can you add to the test in print_ut.c?

Sure, is that a hard-requirement for such a minor patch ? Is there an example for the other %u / %d variants ?

#3 lore raw

PATCH 1/1 tiny-printf: Support %i

Hi Marek,

Quoted 16 lines · Marek Vasut
Quoted 14 lines · Simon Glass

Hi Marek,

Quoted 10 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 4 lines
Quoted 1 line

Can you add to the test in print_ut.c?

Sure, is that a hard-requirement for such a minor patch ? Is there an example for the other %u / %d variants ?

I think we should promote adding tests.

Just add a few lines and then others can contribute more.

The current tests use sprintf(). But there is a ut_check_console_line() that should help. It's really easy to use.

See here for an example:

https://gitlab.denx.de/u-boot/custodians/u-boot-dm/-/blob/corp6-working/test/dm/acpi.c#L262

Regards,
Simon

#4 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 26 lines · Simon Glass

Hi Marek,

Quoted 16 lines · Marek Vasut
Quoted 14 lines · Simon Glass

Hi Marek,

Quoted 10 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
Quoted 4 lines
Quoted 1 line

Can you add to the test in print_ut.c?

Sure, is that a hard-requirement for such a minor patch ? Is there an example for the other %u / %d variants ?

I think we should promote adding tests.

Quoted 8 lines

Just add a few lines and then others can contribute more.

The current tests use sprintf(). But there is a ut_check_console_line() that should help. It's really easy to use.

See here for an example:

https://gitlab.denx.de/u-boot/custodians/u-boot-dm/-/blob/corp6-working/test/dm/acpi.c#L262

These look like ACPI tests, not printf() tests ?

Also, please answer my question -- there are currently no tests for the other variants, so you expect me to implement those as well, correct ?

#5 lore raw

PATCH 1/1 tiny-printf: Support %i

In reply to Marek Vasut
Quoted 23 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 8 lines
lib/tiny-printf.c +1 -1
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem? Thanks!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200413/bd53805c/attachment.sig>
#6 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

#7 lore raw

PATCH 1/1 tiny-printf: Support %i

Hi Marek,

Quoted 43 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                             goto abort;
                     case 'u':
                     case 'd':
+                    case 'i':
                             div = 1000000000;
                             if (islong) {
                                     num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                     num = va_arg(va, unsigned int);
                             }
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-                            if (ch == 'd') {
+                            if (ch != 'u') {
                                     if (islong && (long)num < 0) {
                                             num = -(long)num;
                                             out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

Quoted 6 lines

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

This is not an issue of code size. It is simply that we have a lot of untested code, and we all need to club together to fix that.

Re DM/DT, if you have thoughts on how to improve it, please let me know. I am very concerned about it also.

If some of the most senior maintainers in U-Boot are so opposed to adding tests with new code, how do we expect others to make the effort? Marek, you were one of the most vocal advocates of a longer release cycle because you were seeing lots of breakage that maintainers didn't have time to find. How do I square that with the avoidance of adding tests?

Regards,
Simon

#8 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 51 lines · Simon Glass

Hi Marek,

Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                             goto abort;
                     case 'u':
                     case 'd':
+                    case 'i':
                             div = 1000000000;
                             if (islong) {
                                     num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                     num = va_arg(va, unsigned int);
                             }
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-                            if (ch == 'd') {
+                            if (ch != 'u') {
                                     if (islong && (long)num < 0) {
                                             num = -(long)num;
                                             out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

Quoted 4 lines
Quoted 1 line

The code grows by 6 bytes.

This is not an issue of code size. It is simply that we have a lot of untested code, and we all need to club together to fix that.

It seems to me that Toms counter-argument is code size.

Quoted 2 lines

Re DM/DT, if you have thoughts on how to improve it, please let me know. I am very concerned about it also.

I think I mentioned it before, what about bypassing uclasses which have only one driver instance in them . Then the whole code for tracking instances can be optimized away for that particular uclass, which should save quite a bit of space.

Quoted 6 lines

If some of the most senior maintainers in U-Boot are so opposed to adding tests with new code, how do we expect others to make the effort? Marek, you were one of the most vocal advocates of a longer release cycle because you were seeing lots of breakage that maintainers didn't have time to find. How do I square that with the avoidance of adding tests?

You are wrong, I am not opposed to adding tests. I am opposed to blocking trivial patches which take minutes to implement by requesting adding an entire class of tests, which would likely take days to implement properly. That sheer difference in magnitude does not sound right to me.

#9 lore raw

PATCH 1/1 tiny-printf: Support %i

In reply to Marek Vasut
Quoted 42 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

Quoted 6 lines

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth. So it's just debug prints you were doing that ran in to a problem? Thanks!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200413/3aaccf03/attachment.sig>
#10 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 48 lines · Tom Rini
Quoted 47 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

Quoted 4 lines
Quoted 1 line

The code grows by 6 bytes.

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

#11 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 49 lines · Marek Vasut
Quoted 48 lines · Tom Rini
Quoted 47 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

The code grows by 6 bytes.

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

Quoted 6 lines
Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

Not all 400 of them but the ones that are expected to be used in SPL and with TINY_PRINTF need to, yes. Go look at the git log of tiny-printf.c, we've changed things around to avoid growth when at all possible. Because yes, I don't want to grow a few hundred boards by 6 bytes when we have a reasonable alternative. There's 300 hits, of which a dozen are non-debug and likely to ever be in SPL code. And no, this isn't the first time I've raised such an issue, it's just the first time you've been hit by this, sorry.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200414/6fbbc19a/attachment.sig>
#12 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 52 lines · Tom Rini
Quoted 51 lines · Marek Vasut
Quoted 50 lines · Tom Rini
Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 10 lines
Quoted 6 lines
Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

Not all 400 of them but the ones that are expected to be used in SPL and with TINY_PRINTF need to, yes. Go look at the git log of tiny-printf.c, we've changed things around to avoid growth when at all possible.

I appreciate that. However, I would also appreciate if printf() behaved in a sane manner, and missing %i support is really weird.

Quoted 3 lines

Because yes, I don't want to grow a few hundred boards by 6 bytes when we have a reasonable alternative. There's 300 hits, of which a dozen are non-debug and likely to ever be in SPL code.

Why don't we instead replace %d with %i altogether then ? The %d seems to be seldom used outside of U-Boot, where it is only used because of this tiny-printf limitation, while %i is used quite often.

Quoted 3 lines

And no, this isn't the
first time I've raised such an issue, it's just the first time you've been hit by this, sorry.

Does this therefore set a precedent that we are allowed to block any and all patches which grow SPL size, no matter how useful they might be ?

#13 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 88 lines · Marek Vasut
Quoted 52 lines · Tom Rini
Quoted 51 lines · Marek Vasut
Quoted 50 lines · Tom Rini
Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 10 lines
Quoted 6 lines
Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

Not all 400 of them but the ones that are expected to be used in SPL and with TINY_PRINTF need to, yes. Go look at the git log of tiny-printf.c, we've changed things around to avoid growth when at all possible.

I appreciate that. However, I would also appreciate if printf() behaved in a sane manner, and missing %i support is really weird.

Quoted 3 lines

Because yes, I don't want to grow a few hundred boards by 6 bytes when we have a reasonable alternative. There's 300 hits, of which a dozen are non-debug and likely to ever be in SPL code.

Why don't we instead replace %d with %i altogether then ? The %d seems to be seldom used outside of U-Boot, where it is only used because of this tiny-printf limitation, while %i is used quite often.

Quoted 6 lines
Quoted 3 lines

And no, this isn't the
first time I've raised such an issue, it's just the first time you've been hit by this, sorry.

Does this therefore set a precedent that we are allowed to block any and all patches which grow SPL size, no matter how useful they might be ?

This is following the precedent that was set for tiny printf a while ago with some other "it would be nice if..." format that could instead be handled differently, again for the case of tiny printf. It is not supposed to cover everything, or most things. It is supposed to let SPL/TPL still have printf in otherwise very tight situations.

And as a reminder, I throw every PR through a before/after size check and flag growth that's global and not fixing a bug that can't be fixed some other way. Change your prints to %d and fix the problem without a size change.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200414/ec952317/attachment.sig>
#14 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 93 lines · Tom Rini
Quoted 92 lines · Marek Vasut
Quoted 52 lines · Tom Rini
Quoted 51 lines · Marek Vasut
Quoted 50 lines · Tom Rini
Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 10 lines
Quoted 6 lines
Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

Not all 400 of them but the ones that are expected to be used in SPL and with TINY_PRINTF need to, yes. Go look at the git log of tiny-printf.c, we've changed things around to avoid growth when at all possible.

I appreciate that. However, I would also appreciate if printf() behaved in a sane manner, and missing %i support is really weird.

Quoted 3 lines

Because yes, I don't want to grow a few hundred boards by 6 bytes when we have a reasonable alternative. There's 300 hits, of which a dozen are non-debug and likely to ever be in SPL code.

Why don't we instead replace %d with %i altogether then ? The %d seems to be seldom used outside of U-Boot, where it is only used because of this tiny-printf limitation, while %i is used quite often.

Quoted 3 lines

And no, this isn't the
first time I've raised such an issue, it's just the first time you've been hit by this, sorry.

Quoted 8 lines
Quoted 2 lines

Does this therefore set a precedent that we are allowed to block any and all patches which grow SPL size, no matter how useful they might be ?

This is following the precedent that was set for tiny printf a while ago with some other "it would be nice if..." format that could instead be handled differently, again for the case of tiny printf. It is not supposed to cover everything, or most things. It is supposed to let SPL/TPL still have printf in otherwise very tight situations.

Except the way it is right now, a lot of output is broken in SPL in an inobvious way, which makes working with and/or debugging SPL difficult and special, compared to U-Boot and other software.

Quoted 4 lines

And as a reminder, I throw every PR through a before/after size check and flag growth that's global and not fixing a bug that can't be fixed some other way. Change your prints to %d and fix the problem without a size change.

Sorry, no, "change your formatting strings to cater for our broken printf() implementation" really is not the solution here.

#15 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 103 lines · Marek Vasut
Quoted 93 lines · Tom Rini
Quoted 92 lines · Marek Vasut
Quoted 52 lines · Tom Rini
Quoted 51 lines · Marek Vasut
Quoted 50 lines · Tom Rini
Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 10 lines
Quoted 6 lines
Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

Not all 400 of them but the ones that are expected to be used in SPL and with TINY_PRINTF need to, yes. Go look at the git log of tiny-printf.c, we've changed things around to avoid growth when at all possible.

I appreciate that. However, I would also appreciate if printf() behaved in a sane manner, and missing %i support is really weird.

Quoted 3 lines

Because yes, I don't want to grow a few hundred boards by 6 bytes when we have a reasonable alternative. There's 300 hits, of which a dozen are non-debug and likely to ever be in SPL code.

Why don't we instead replace %d with %i altogether then ? The %d seems to be seldom used outside of U-Boot, where it is only used because of this tiny-printf limitation, while %i is used quite often.

Quoted 3 lines

And no, this isn't the
first time I've raised such an issue, it's just the first time you've been hit by this, sorry.

Quoted 8 lines
Quoted 2 lines

Does this therefore set a precedent that we are allowed to block any and all patches which grow SPL size, no matter how useful they might be ?

This is following the precedent that was set for tiny printf a while ago with some other "it would be nice if..." format that could instead be handled differently, again for the case of tiny printf. It is not supposed to cover everything, or most things. It is supposed to let SPL/TPL still have printf in otherwise very tight situations.

Quoted 3 lines

Except the way it is right now, a lot of output is broken in SPL in an inobvious way, which makes working with and/or debugging SPL difficult and special, compared to U-Boot and other software.

No, it's not "a lot", it's whatever you ran in to. We don't have a lot of "%i" usage to start with and even less that's not a debug print and also in SPL code.

Quoted 7 lines
Quoted 4 lines

And as a reminder, I throw every PR through a before/after size check and flag growth that's global and not fixing a bug that can't be fixed some other way. Change your prints to %d and fix the problem without a size change.

Sorry, no, "change your formatting strings to cater for our broken printf() implementation" really is not the solution here.

Except it is, for the specifically limited printf limitation. Adapt the print for the limitation and move on. It's not even "rework the code slightly so the print is correct", it's switch to %d.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200414/e7a123e8/attachment.sig>
#16 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 104 lines · Tom Rini
Quoted 103 lines · Marek Vasut
Quoted 101 lines · Tom Rini
Quoted 88 lines · Marek Vasut
Quoted 52 lines · Tom Rini
Quoted 51 lines · Marek Vasut
Quoted 50 lines · Tom Rini
Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 				goto abort;
 			case 'u':
 			case 'd':
+			case 'i':
 				div = 1000000000;
 				if (islong) {
 					num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
 					num = va_arg(va, unsigned int);
 				}
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-				if (ch == 'd') {
+				if (ch != 'u') {
 					if (islong && (long)num < 0) {
 						num = -(long)num;
 						out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 10 lines
Quoted 6 lines
Quoted 2 lines

So it's just debug prints you were doing that ran in to a problem? Thanks!

git grep %i indicates ~400 sites where %i is used, so no, not just debug prints. All of those are broken. And no, I'm not inclined to patch all the code to use %d instead of %i just because handling %i is a problem.

Not all 400 of them but the ones that are expected to be used in SPL and with TINY_PRINTF need to, yes. Go look at the git log of tiny-printf.c, we've changed things around to avoid growth when at all possible.

I appreciate that. However, I would also appreciate if printf() behaved in a sane manner, and missing %i support is really weird.

Quoted 3 lines

Because yes, I don't want to grow a few hundred boards by 6 bytes when we have a reasonable alternative. There's 300 hits, of which a dozen are non-debug and likely to ever be in SPL code.

Why don't we instead replace %d with %i altogether then ? The %d seems to be seldom used outside of U-Boot, where it is only used because of this tiny-printf limitation, while %i is used quite often.

Quoted 6 lines
Quoted 3 lines

And no, this isn't the
first time I've raised such an issue, it's just the first time you've been hit by this, sorry.

Does this therefore set a precedent that we are allowed to block any and all patches which grow SPL size, no matter how useful they might be ?

This is following the precedent that was set for tiny printf a while ago with some other "it would be nice if..." format that could instead be handled differently, again for the case of tiny printf. It is not supposed to cover everything, or most things. It is supposed to let SPL/TPL still have printf in otherwise very tight situations.

Quoted 7 lines
Quoted 3 lines

Except the way it is right now, a lot of output is broken in SPL in an inobvious way, which makes working with and/or debugging SPL difficult and special, compared to U-Boot and other software.

No, it's not "a lot", it's whatever you ran in to. We don't have a lot of "%i" usage to start with and even less that's not a debug print and also in SPL code.

And it prevents any sort of sane code reuse, unless you want to spam the code with #ifdef TINY_PRINTF.

Quoted 11 lines
Quoted 7 lines
Quoted 4 lines

And as a reminder, I throw every PR through a before/after size check and flag growth that's global and not fixing a bug that can't be fixed some other way. Change your prints to %d and fix the problem without a size change.

Sorry, no, "change your formatting strings to cater for our broken printf() implementation" really is not the solution here.

Except it is, for the specifically limited printf limitation. Adapt the print for the limitation and move on. It's not even "rework the code slightly so the print is correct", it's switch to %d.

See above.

#17 lore raw

PATCH 1/1 tiny-printf: Support %i

In reply to Marek Vasut

Hi Marek,

Quoted 60 lines · Marek Vasut
Quoted 58 lines · Tom Rini
Quoted 56 lines · Marek Vasut
Quoted 48 lines · Tom Rini
Quoted 47 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                 goto abort;
                         case 'u':
                         case 'd':
+                        case 'i':
                                 div = 1000000000;
                                 if (islong) {
                                         num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                         num = va_arg(va, unsigned int);
                                 }
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-                                if (ch == 'd') {
+                                if (ch != 'u') {
                                         if (islong && (long)num < 0) {
                                                 num = -(long)num;
                                                 out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

Quoted 4 lines
Quoted 1 line

The code grows by 6 bytes.

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

Quoted 6 lines
Quoted 2 lines

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

I think this is a good point. For serial we have a debug UART. I am think it would be useful to have a 'simple bypass' for more subsystems. For example, for I2C we could have a simple option to access a single I2C driver directly, bypassing driver model. Of course this is painful to do before we have completed I2C migration.

[..]

Regards,
Simon

#18 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 70 lines · Simon Glass

Hi Marek,

Quoted 60 lines · Marek Vasut
Quoted 58 lines · Tom Rini
Quoted 56 lines · Marek Vasut
Quoted 48 lines · Tom Rini
Quoted 47 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                 goto abort;
                         case 'u':
                         case 'd':
+                        case 'i':
                                 div = 1000000000;
                                 if (islong) {
                                         num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                         num = va_arg(va, unsigned int);
                                 }
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-                                if (ch == 'd') {
+                                if (ch != 'u') {
                                         if (islong && (long)num < 0) {
                                                 num = -(long)num;
                                                 out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

Quoted 4 lines
Quoted 1 line

The code grows by 6 bytes.

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

Quoted 6 lines
Quoted 2 lines

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 5 lines

I think this is a good point. For serial we have a debug UART. I am think it would be useful to have a 'simple bypass' for more subsystems. For example, for I2C we could have a simple option to access a single I2C driver directly, bypassing driver model. Of course this is painful to do before we have completed I2C migration.

Right. Although it could be prototyped e.g. on the UART subssystem, which is a good candidate.

#19 lore raw

PATCH 1/1 tiny-printf: Support %i

Hi Marek,

Quoted 72 lines · Marek Vasut
Quoted 70 lines · Simon Glass

Hi Marek,

Quoted 66 lines · Marek Vasut
Quoted 52 lines · Tom Rini
Quoted 51 lines · Marek Vasut
Quoted 50 lines · Tom Rini
Quoted 48 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                 goto abort;
                         case 'u':
                         case 'd':
+                        case 'i':
                                 div = 1000000000;
                                 if (islong) {
                                         num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                         num = va_arg(va, unsigned int);
                                 }
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-                                if (ch == 'd') {
+                                if (ch != 'u') {
                                         if (islong && (long)num < 0) {
                                                 num = -(long)num;
                                                 out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

The code grows by 6 bytes.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

Quoted 8 lines
Quoted 5 lines

I think this is a good point. For serial we have a debug UART. I am think it would be useful to have a 'simple bypass' for more subsystems. For example, for I2C we could have a simple option to access a single I2C driver directly, bypassing driver model. Of course this is painful to do before we have completed I2C migration.

Right. Although it could be prototyped e.g. on the UART subssystem, which is a good candidate.

Yes it is on my mind, once i get my lab working properly. Will have a think about it...

Regards,
Simon

#20 lore raw

PATCH 1/1 tiny-printf: Support %i

Quoted 1 line · Simon Glass

Hi Marek,

Hi,

Quoted 79 lines · Marek Vasut
Quoted 78 lines · Simon Glass
Quoted 75 lines

Hi Marek,

Quoted 60 lines · Marek Vasut
Quoted 58 lines · Tom Rini
Quoted 56 lines · Marek Vasut
Quoted 48 lines · Tom Rini
Quoted 47 lines · Marek Vasut
Quoted 28 lines · Tom Rini
Quoted 26 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>
---
 lib/tiny-printf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
lib/tiny-printf.c +1 -0
index 1138c7012a..8fc7e48d99 100644
@@ -242,6 +242,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                 goto abort;
                         case 'u':
                         case 'd':
+                        case 'i':
                                 div = 1000000000;
                                 if (islong) {
                                         num = va_arg(va, unsigned long);
@@ -251,7 +252,7 @@ static int _vprintf(struct printf_info *info, const char *fmt, va_list va)
                                         num = va_arg(va, unsigned int);
                                 }
 
Quoted 7 lines

lib/tiny-printf.c static int _vprintf(struct printf_info *info, const char *fmt, va_list va) · line 258

Quoted 5 lines
lib/tiny-printf.c +1 -1
-                                if (ch == 'd') {
+                                if (ch != 'u') {
                                         if (islong && (long)num < 0) {
                                                 num = -(long)num;
                                                 out(info, '-');

How much does the size change and where do we see this as a problem?

Any code which uses %i in SPL just misbehaves, e.g. printf("%s[%i] value=%x", __func__, __LINE__, val);
prints function name and then incorrect value, because %i is ignored. This is also documented in the commit message.

U-Boot grows in size massively due to all the DM/DT bloat which is being forced upon everyone, but there the uncontrolled growth is apparently OK even if it brings no obvious improvement, rather the opposite. And yet here, size increase suddenly matters? Sorry, that's not right.

Quoted 4 lines
Quoted 1 line

The code grows by 6 bytes.

Yes, it matters for _tiny-printf_ as that's where we have little to no room for growth.

How many systems that use tiny-printf in SPL are also forced to use DM in SPL ?

Quoted 6 lines
Quoted 2 lines

I don't know how many times I've said no one is forced to switch to DM in SPL.

This is beside the point, there are boards which use SPL and DM, because the non-DM drivers are steadily going away. So the growth in SPL size is there, either directly or as a side-effect.

I think this is a good point. For serial we have a debug UART. I am think it would be useful to have a 'simple bypass' for more subsystems. For example, for I2C we could have a simple option to access a single I2C driver directly, bypassing driver model. Of course this is painful to do before we have completed I2C migration.

Quoted 5 lines
Quoted 2 lines

Right. Although it could be prototyped e.g. on the UART subssystem, which is a good candidate.

Yes it is on my mind, once i get my lab working properly. Will have a think about it...

Thanks. This is long overdue to be implemented.

#21 lore raw

PATCH 1/1 tiny-printf: Support %i

In reply to Marek Vasut
Quoted 6 lines · Marek Vasut

The most basic printf("%i", value) formating string was missing, add it for the sake of convenience.

Cc Simon Glass <[email protected]>
Cc Stefan Roese <[email protected]>

Applied to u-boot/master, thanks!

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200501/807900a6/attachment.sig>