`boundary = 0` is ignored in 4.0.0's geom_histogram()

[dkahle]

The move from 3.5.2 to 4.0.0 seems to introduce a bug that ignores boundary settings in geom_histogram(). Reprexes below.

---

Reprex using CRAN's 4.0.0

library("ggplot2")
sessionInfo()$otherPkgs$ggplot2$Version
#> [1] "4.0.0"

set.seed(1234)
df <- data.frame(x = abs(rnorm(1e4)))

# histogram of exponential data, notice first bin crosses 0 (all data is > 0)
ggplot(df, aes(x)) +
  geom_histogram(bins = 101)

# setting boundary = 0, but no change
ggplot(df, aes(x)) +
  geom_histogram(bins = 101, boundary = 0)

# zooming in on 0-boundary crossover
ggplot(df, aes(x)) +
  geom_histogram(bins = 101, boundary = 0) +
  coord_cartesian(xlim = c(-.2, .5))

---

Reprex using CRAN's 3.5.2

library("ggplot2")
sessionInfo()$otherPkgs$ggplot2$Version
#> [1] "3.5.2"

set.seed(1234)
df <- data.frame(x = abs(rnorm(1e4)))

# histogram of exponential data, notice first bin crosses 0 (all data is > 0)
ggplot(df, aes(x)) +
  geom_histogram(bins = 101)

# setting boundary = 0, which *does* show change
ggplot(df, aes(x)) +
  geom_histogram(bins = 101, boundary = 0)

# zooming in on 0-boundary crossover
ggplot(df, aes(x)) +
  geom_histogram(bins = 101, boundary = 0) +
  coord_cartesian(xlim = c(-.2, .5))

[teunbrand]

Thanks for the report! I can confirm this on my end.
This is likely a consequence of #5891.

Down in the internals, I think the following is going wrong.

ggplot2:::bin_breaks_bins(x_range = c(0, 10), bins = 2, boundary = 0)
#> <Bins>
#> * [-5, 5]
#> * ( 5,15]

^{Created on 2025-10-06 with reprex v2.1.1}