develop/ktcc: libc_scanf_add_double_parse #407

2026-04-05T20:39:42Z

DVZverrr commented

2026-04-05 20:39:42 +00:00

First-time contributor

Добавлена возможность считывать float и double
ex:
scanf("%f", &float_val);
scanf("%lf", &double_val);

Попутно починен баг в функции itoa(...), который проявлялся при запуске libc_test из набора примеров.

Добавлена возможность считывать float и double ex: scanf("%f", &float_val); scanf("%lf", &double_val); Попутно починен баг в функции itoa(...), который проявлялся при запуске libc_test из набора примеров.

DVZverrr force-pushed libc_scanf_add_double_parse from 82fa1f7525 to f55d621fb7

2026-04-05 20:43:11 +00:00

Compare

DVZverrr changed title from ~~libc_scanf_add_double_parse~~ to develop/ktcc: libc_scanf_add_double_parse

2026-04-19 13:49:21 +00:00

ace-dent commented

2026-04-20 11:26:50 +00:00

@DVZverrr - Please note that the file path for ktcc was recently changed with #425. This will conflict with your changes.

@DVZverrr - Please note that the file path for `ktcc` was recently changed with #425. This will conflict with your changes.

DVZverrr added 3 commits 2026-04-21 22:06:42 +00:00

Add parsing floats and doubles with scanf 7ea383be7a

fix bug in itoa function 9787d27790

string.h теперь не нужен

Build system / Check kernel codestyle (pull_request) Successful in 2m41s

Details

Build system / Build (pull_request) Successful in 11m24s

Details

db3e211f81

DVZverrr force-pushed libc_scanf_add_double_parse from f55d621fb7 to db3e211f81

2026-04-21 22:06:42 +00:00

Compare

DVZverrr commented

2026-04-21 22:17:28 +00:00

First-time contributor

Rebased to fresh main

Burer approved these changes 2026-04-25 10:32:02 +00:00

Dismissed

Burer left a comment

Hi, @DVZverrr!

I reviewed this PR using AI and found a few problems that seems worth fixing before this PR will be merged. Not 100% sure about all of them, so please, check and provide some feedback.

Bugs:

%f ignores field width

File: programs/develop/ktcc/libc.obj/source/stdio/vsscanf.c

The new floating-point path calls strtod(q, ...) directly, so a format like %3f consumes the full input number instead of at most 3 characters.

Example:
```
sscanf("1234", "%3f%d", &f, &i);
```
Expected behavior: parse 123 as the float and 4 as the integer.
Current behavior: %f consumes 1234, so the following %d fails.
%Lf stores into the wrong destination type

File: programs/develop/ktcc/libc.obj/source/stdio/vsscanf.c

The L modifier sets rank = rank_longlong, but the new code treats any rank > rank_int as double *.

For %Lf, the caller provides a long double *, not a double *. If long double is wider than double in this toolchain, this writes an incomplete/corrupt value.

Improvements:

Only %f is implemented

File: programs/develop/ktcc/libc.obj/source/stdio/vsscanf.c

Standard scanf floating-point conversions also include %e, %E, %g, and %G. Since strtod() already handles exponent notation, this should mostly be handled by dispatching those specifiers to the same floating-point parsing path.
Floating-point width handling needs an explicit bounded input path

Unlike strntoumax(), strtod() does not receive a width argument here. The implementation should limit the input manually when a field width is specified, for example by parsing from a temporary bounded buffer.
The floating-point conversion dispatch should be broadened and clarified

Current code only has:
```
case 'f':
```
It should probably become something like:
```
case 'e':
case 'E':
case 'f':
case 'F':
case 'g':
case 'G':
```
Formatting of the new block does not match the surrounding style

The indentation around if(rank > rank_int) and the va_arg() assignments is inconsistent with the rest of vsscanf.c.
%a / %A should be considered explicitly

If this libc’s strtod() does not support hexadecimal floating-point input, it is fine to leave %a / %A unsupported in this PR. But that limitation should be intentional.

Hi, @DVZverrr! I reviewed this PR using AI and found a few problems that seems worth fixing before this PR will be merged. Not 100% sure about all of them, so please, check and provide some feedback. Bugs: - [ ] `%f` ignores field width File: `programs/develop/ktcc/libc.obj/source/stdio/vsscanf.c` The new floating-point path calls `strtod(q, ...)` directly, so a format like `%3f` consumes the full input number instead of at most 3 characters. Example: ```c sscanf("1234", "%3f%d", &f, &i); ``` Expected behavior: parse `123` as the float and `4` as the integer. Current behavior: `%f` consumes `1234`, so the following `%d` fails. - [ ] `%Lf` stores into the wrong destination type File: `programs/develop/ktcc/libc.obj/source/stdio/vsscanf.c` The `L` modifier sets `rank = rank_longlong`, but the new code treats any `rank > rank_int` as `double *`. For `%Lf`, the caller provides a `long double *`, not a `double *`. If `long double` is wider than `double` in this toolchain, this writes an incomplete/corrupt value. Improvements: - [ ] Only `%f` is implemented File: `programs/develop/ktcc/libc.obj/source/stdio/vsscanf.c` Standard `scanf` floating-point conversions also include `%e`, `%E`, `%g`, and `%G`. Since `strtod()` already handles exponent notation, this should mostly be handled by dispatching those specifiers to the same floating-point parsing path. - [ ] Floating-point width handling needs an explicit bounded input path Unlike `strntoumax()`, `strtod()` does not receive a width argument here. The implementation should limit the input manually when a field width is specified, for example by parsing from a temporary bounded buffer. - [ ] The floating-point conversion dispatch should be broadened and clarified Current code only has: ```c case 'f': ``` It should probably become something like: ```c case 'e': case 'E': case 'f': case 'F': case 'g': case 'G': ``` - [ ] Formatting of the new block does not match the surrounding style The indentation around `if(rank > rank_int)` and the `va_arg()` assignments is inconsistent with the rest of `vsscanf.c`. - [ ] `%a` / `%A` should be considered explicitly If this libc’s `strtod()` does not support hexadecimal floating-point input, it is fine to leave `%a` / `%A` unsupported in this PR. But that limitation should be intentional.

Burer dismissed Burer's review 2026-04-26 04:45:07 +00:00

Burer requested changes 2026-04-26 04:46:07 +00:00

Burer left a comment

To previous.

DVZverrr commented

2026-04-29 09:14:25 +00:00

First-time contributor

In progress

DVZverrr added 1 commit 2026-04-30 00:11:51 +00:00

vsscanf: width for float added; %g %e added; libc_test updated

Build system / Check kernel codestyle (pull_request) Successful in 49s

Details

Build system / Build (pull_request) Successful in 10m37s

Details

3781e66285

strtof added to libc
strtold added to libc

DVZverrr commented

2026-04-30 00:28:52 +00:00

First-time contributor

Support for %e %g is added
%G and %E are not relevant for scanf. These options are used for printf
Support for width for %f %e %g is added
Type mapping now:
%f -> float
%lf -> double
%Lf -> long double

Some related assertions have been added to libc_test.c as tests

P.S.
I've left %a and %A unsupported. I think it would be better to implement these options when needed.

Support for %e %g is added %G and %E are not relevant for scanf. These options are used for printf Support for width for %f %e %g is added Type mapping now: %f -> float %lf -> double %Lf -> long double Some related assertions have been added to libc_test.c as tests P.S. I've left %a and %A unsupported. I think it would be better to implement these options when needed.

Build system / Check kernel codestyle (pull_request) Successful in 49s

Details

Build system / Build (pull_request) Successful in 10m37s

Details

Checking for merge conflicts…

View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u http://git.kolibrios.org/DVZverrr/kolibrios libc_scanf_add_double_parse:DVZverrr-libc_scanf_add_double_parse

git checkout DVZverrr-libc_scanf_add_double_parse

Sign in to join this conversation.

3 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: KolibriOS/kolibrios#407

Allow edits from maintainers